<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:dc="http://purl.org/dc/elements/1.1/">
<channel>
<title>On The Edge — the Preppr blog</title>
<link>https://blog.preppr.ai/</link>
<description>Essays, product news, and field notes from Preppr on AI, emergency management, and building a preparedness program that compounds instead of resetting to zero.</description>
<language>en-us</language>
<lastBuildDate>Fri, 20 Nov 2026 12:00:00 GMT</lastBuildDate>
<atom:link href="https://blog.preppr.ai/feed.xml" rel="self" type="application/rss+xml"/>
<item>
<title>Preparedness That Keeps Up With Reality</title>
<link>https://blog.preppr.ai/a/preparedness-that-keeps-up-with-reality/</link>
<guid isPermaLink="true">https://blog.preppr.ai/a/preparedness-that-keeps-up-with-reality/</guid>
<pubDate>Fri, 20 Nov 2026 12:00:00 GMT</pubDate>
<dc:creator>Justin Snair</dc:creator>
<category>Industry Thought</category>
<description>Practice is the flywheel of preparedness. Preppr makes it fast, and keeps it grounded in what’s true about your place.</description>
<content:encoded><![CDATA[<p>What kept bothering me is that preparedness is supposed to be continuous: a living thing that keeps up with a changing place. Hazards shift, capabilities change, and people move on. A plan written for last year’s reality ages on the shelf the day it’s approved.</p><p>What keeps preparedness alive is practice, meaning exercises and simulations. Practice is how a team finds out whether the plan still matches the place, and it’s the flywheel that turns findings into better plans. But that flywheel has always turned slowly. Exercises and simulations were slow to design and slow to run, and by the time you’d designed and run one, the reality had moved.</p><h2 id="speed-was-the-first-problem">Speed was the first problem</h2><p>So the first problem was speed: design and run exercises and simulations fast enough to keep up with the place they’re meant to test.</p><p>We solved it. Preppr designs and runs exercises and simulations in a fraction of the time they used to take. Producing the content isn’t the hard part anymore.</p><p>That shift changes what a program can do. When an exercise no longer takes months of staff time, you can practice more often, practice more of the plan, and come back to the same scenario after you’ve fixed what it found. Speed is what makes practice continuous.</p><h2 id="the-harder-problem-is-context">The harder problem is context</h2><p>Speed on its own isn’t enough. AI is only as useful as what it knows about your place. An exercise built from the internet’s average disaster tests a generic jurisdiction, not yours. It can be fluent and polished and still be wrong about your hospitals, your shelters, your agreements and your people.</p><p>So the deeper problem is context: capturing what’s true about your place, keeping it accurate, finding the right piece at the right moment, and maintaining it as things change. When AI designs or runs an exercise, it should be working from your reality, not a generic one.</p><p>That’s what we’ve built Preppr around: a living context layer. It goes well beyond passages pulled from documents. Preppr observes what happens in the room, finds patterns, forms rules and confirms them with you, relates everything it knows, surfaces gaps and tracks the improvements that close them. And it checks its own work. It comes down to five jobs.</p><h3>Capture</h3><p>Exercises start from your own plans, risk data, after-action reports and past findings, not the internet’s average disaster. Preppr reads more than the text: the diagrams, org charts and tables in your plans are read too.</p><p>Documents are only part of the record. The work your team does in Preppr goes into it as well: exercise designs and the conversations that shaped them, the findings from every exercise you run, and the after-action record of every simulation. Preppr’s Observer can also join your meetings and capture the decisions, action items and moments people flag. A meeting tied to an exercise joins the record by itself, and other meetings wait for a person to review them.</p><h3>Accuracy</h3><p>Preppr pulls values from your documents with the page and sentence they came from, so anyone can check where a number or a commitment was found. Each value stays a candidate until a person checks it. When two plans give different values for the same thing, Preppr flags the disagreement and shows both sources, and you can aim an exercise right at that seam to see which plan people follow. When you upload what looks like a new version of a plan, Preppr suggests the match and a person confirms it. The old version then drops out of search, and any answer that cited it is marked as made against an earlier version. Every Ask Preppr answer is labeled by how well your sources support it: grounded, needs review, or largely unverifiable. You can see at a glance what rests on your own sources and what still needs a person to look at it.</p><h3>Connection</h3><p>Preppr doesn’t just store what it reads. It forms rules from the “shall” and “must” in your own plans. And because your team’s work flows into the same record, Preppr can see how your team actually works, not just what the plan says. When the same issue comes up on separate occasions, such as a finding from an exercise, a decision made while designing another and an action item from a planning meeting, Preppr draws out the pattern and sets it against your rules: does practice match the plan, or drift from it?</p><p>The bar for a pattern is deliberately high. Points from a single conversation count once, however often they come up, and a simulation played only by AI crews can support a pattern but never create one on its own. Rules and patterns stay suggestions until a person accepts them. Preppr also learns your team’s preferences from how people correct and edit its work, and it lists every one it has picked up.</p><p>It relates plans, requirements, facts and findings on one graph, each traced to its source, so anything Preppr knows can be followed back to where it came from.</p><h3>Retrieval</h3><p>Knowing a lot isn’t the same as using the right thing at the right time. When Preppr designs an exercise, it searches the sources attached to that exercise and the collections you choose. Ask Preppr answers from the documents you select and your knowledge base, and it cites the passages it used. In a simulation, you decide how well each seat can find things in your plans, from perfectly to not at all, so you can test what happens when people can’t find the procedure.</p><h3>Maintenance</h3><p>Context goes stale the same way plans do. Findings from an exercise can become improvement work, each with an owner and a due date, and Preppr reminds owners as dates come due. When a new document looks like it closes a gap, Preppr suggests it to the owner, but it never closes a gap by itself. You can design the next exercise straight from the open actions, so the fix gets tested.</p><p>And Preppr reviews itself. Every night it goes over what it has learned. It makes small changes on its own, like new topics and connections, but only after testing that they don’t make it worse at finding your sources, and every one can be undone. Bigger changes, like revising a rule or settling a clash between two rules, wait for a person to decide.</p><p>Together, that’s a living context layer: a record of your place that grows with every plan, meeting, exercise and fix, and keeps up as the place changes.</p><h2 id="the-judgment-stays-with-your-team">The judgment stays with your team</h2><p>Grounding goes beyond your own documents. Scenarios are checked against FEMA and Hazus data for your jurisdiction, so the hazard in an exercise behaves the way it would in your place.</p><p>And AI findings stay evidence until a person confirms them. Preppr can surface a gap, cite where it came from and suggest what to do about it, but your team decides what’s true and what changes. The point isn’t to hand preparedness to a machine. It’s to give the people responsible for it a faster, better-grounded way to practice.</p><h2 id="keeping-up">Keeping up</h2><p>Preparedness that keeps up with reality needs both halves: practice fast enough to run often, and a living context layer that keeps the practice about your place. Speed was the first problem. Context is the harder one, and it’s the problem we built Preppr to solve.</p><p><a href="https://preppr.ai/how-it-works">See how Preppr works</a> or <a href="https://preppr.ai/demo">talk to us</a>.</p>]]></content:encoded>
</item>
<item>
<title>Practice What Matters: Are We Exercising the Right Disruptions?</title>
<link>https://blog.preppr.ai/a/practice-what-matters/</link>
<guid isPermaLink="true">https://blog.preppr.ai/a/practice-what-matters/</guid>
<pubDate>Thu, 12 Nov 2026 12:00:00 GMT</pubDate>
<dc:creator>Justin Snair</dc:creator>
<category>Industry Thought</category>
<description>Organizations are getting better at practicing for disruption. The next opportunity is making sure that practice is focused on what matters most.</description>
<content:encoded><![CDATA[<p>Every year, organizations invest considerable effort in preparedness. They update plans, run tabletop exercises, simulate incidents, and train teams to respond when things go wrong.</p><p>Much of that work is guided by familiar considerations: regulatory expectations, previous incidents, emerging threats, and the scenarios that teams believe are most likely to occur. These are all reasonable inputs, but they don't necessarily tell us which disruptions would have the greatest consequences.</p><p>There's another question worth asking before deciding what to practice: What would it actually cost the organization if this went wrong?</p><h2 id="the-economics-of-choosing-what-to-practice">The economics of choosing what to practice</h2><p>Imagine an organization planning its next three exercises. One scenario involves a cyberattack, another focuses on a facility outage, and the third examines the loss of a critical supplier.</p><p>Each is plausible. Each presents operational challenges and could reveal important gaps in preparedness. But they may have very different consequences for the business.</p><p>A facility outage might be absorbed through alternative capacity. The loss of a seemingly minor supplier could halt production across multiple locations. A technology disruption might affect several critical processes simultaneously.</p><p>Without understanding those dependencies and their economic consequences, organizations risk allocating the same exercise effort to scenarios with vastly different levels of exposure.</p><p>Preparedness resources are finite. Teams have limited time, budgets, and opportunities to practice together, making the choice of what to exercise a consequential risk management decision.</p><h2 id="exercise-design-is-an-investment-decision">Exercise design is an investment decision</h2><p>The scenarios an organization chooses determine which capabilities receive attention, which assumptions are challenged, and which weaknesses are likely to be discovered.</p><p>Practice a ransomware attack, and teams develop experience coordinating a response to a cyber incident. Practice the loss of a critical production dependency, and teams confront questions about alternative capacity, supplier relationships, operational tolerances, and recovery decisions.</p><p>Both may be valuable. Their relative importance depends on the organization, its operating model, and the consequences of failure.</p><p>The implications extend beyond the exercise itself. Findings influence how teams refine procedures, where they build redundancy, what they escalate to leadership, and which capabilities receive further investment.</p><p>This makes exercise selection an important part of resource allocation. A more economically informed approach could help organizations direct scarce preparedness resources toward the capabilities that protect the greatest value.</p><h2 id="what-happens-when-we-connect-the-numbers-to-the-practice">What happens when we connect the numbers to the practice?</h2><p>Consider a manufacturer that discovers a relatively small supplier supports a component used across several high-value production lines. The supplier's annual contract value may be modest, but the financial consequences of losing it could be substantial.</p><p>That insight changes the exercise.</p><p>Instead of running a general supplier disruption scenario, the organization can practice the specific decisions it would face if that dependency became unavailable. How quickly could alternative sources be activated? What production capacity could be maintained? Which customers would be affected, and who would have authority to make the necessary trade-offs?</p><p>The exercise becomes more meaningful because the organization understands why those decisions matter. It can explore how operational choices affect the continuity of critical activities and identify where existing capabilities may be insufficient.</p><p>The financial exposure also gives leadership a stronger basis for evaluating the findings. Recommendations for additional supplier capacity, revised contracts, or alternative operating arrangements can be considered in the context of the potential consequences they are intended to reduce.</p><p>This is where economic understanding and practical exercises begin to reinforce one another.</p><h2 id="measuring-the-value-of-preparedness">Measuring the value of preparedness</h2><p>Preparedness teams have traditionally relied on measures such as exercise completion, participation, plan coverage, and corrective actions. These provide useful information about activity and progress, but they offer a limited view of how prepared an organization is for its most consequential disruptions.</p><p>Connecting exercises to operational and financial exposure creates additional opportunities to demonstrate value.</p><p>Organizations can examine whether teams are able to maintain critical activities, how quickly decisions can be made, whether recovery strategies perform as expected, and which capability gaps could contribute to greater losses.</p><p>Exercise results alone cannot establish how much money an organization would save during a real incident. They can, however, provide evidence to support better decisions about where improvement is needed and which investments warrant priority.</p><p>Over time, this creates a more useful feedback loop between business exposure, exercise findings, and preparedness investment.</p><h2 id="bringing-economics-and-practice-together">Bringing economics and practice together</h2><p>This thinking is behind Preppr's new collaboration with MEO Continuity.</p><p>MEO helps organizations understand the financial consequences of disruption and identify the operational dependencies that matter most. Preppr enables teams to put their assumptions, plans, and capabilities to the test through AI-powered exercises and simulations.</p><p>The opportunity is to bring greater economic relevance into exercise design and use what organizations learn through practice to inform future decisions.</p><p>For preparedness professionals, that means a stronger basis for deciding what to exercise and explaining why it deserves attention. For business leaders, it creates a clearer connection between the risks facing the organization and the capabilities being developed to manage them.</p><h2 id="practice-what-matters">Practice what matters</h2><p>Every exercise represents an investment in organizational capability. The scenarios we select shape what people learn, which weaknesses we discover, and where improvement efforts are directed.</p><p>As organizations look to strengthen preparedness, there is an opportunity to become more deliberate about those choices. Understanding the economic consequences of disruption can help teams focus their practice on the decisions and dependencies that could make the greatest difference.</p><p>The better we understand what matters to the business, the better we can decide what matters to practice.</p><p><a href="https://preppr.ai/partners/meo-continuity">Learn more about the collaboration between Preppr and MEO Continuity.</a></p>]]></content:encoded>
</item>
<item>
<title>Practice With a Clock, Video With a Label, and a Loop You Can See</title>
<link>https://blog.preppr.ai/a/simulate-video-and-the-loop/</link>
<guid isPermaLink="true">https://blog.preppr.ai/a/simulate-video-and-the-loop/</guid>
<pubDate>Fri, 30 Oct 2026 12:00:00 GMT</pubDate>
<dc:creator>Justin Snair</dc:creator>
<category>Product Update</category>
<description>The Simulate products, generated video I was on the fence about until I wasn't, the two pages that show whether your preparedness loop is actually running, and a free, public set of our skills for whatever AI model you already use. What shipped this month, and why each piece is built the way it is.</description>
<content:encoded><![CDATA[<p>Two weeks ago I wrote several thousand words about procurement being the hard part. While I was writing them, the team shipped more than I could keep track of. We put up a new preppr.ai so the website finally describes the product we have. This post is about what's behind it: the Simulate products, generated video, the two pages that show whether your preparedness loop is actually running, and the skills we've started giving away.</p><p>Usual disclosure. I run the company. Read accordingly.</p><h2 id="simulate-practice-with-a-clock-on-it">Simulate: practice with a clock on it</h2><p>In the last post I said a tabletop tests a plan but doesn't test time to decision, that Europe is well ahead of us on actual simulation, and that the United States is still write a document and talk at each other. I also said the pull toward live response is universal, and that AI hasn't earned response yet. The simulation work is my answer to both of those at once. If people want reps under time pressure, give them reps under time pressure, in a synthetic incident, with the labor done by machines and the decisions kept by people.</p><p>There are four ways to do it now, three of them under a Simulate entry in the sidebar and one inside the exercise designer. Every one of them ends in Improve.</p><p><strong>Simulate Ops</strong> is a multiplayer practice Operation Center. People take seats on a room map. AI crews hold the seats nobody filled. A Threat team plays the hazard and the adversary. Preppr Verified physics is the truth underneath, an incident clock runs (it says hour five of twenty-four, never tick), and everything goes on one ledger. Pick an operating model, Incident Command, business continuity, or one of two FEMA EOC charts, set the dials, and start. The opening situation lands as a full-screen alert and ends on a button that says take your desk. From there your desk sorts the work into decide now, escalated to you, and for awareness. Callers ring your seat, and if the host allows it they ring your actual phone. Your section's AI crews send you one staff memo with a recommendation, the risk and who agreed, instead of a pile of escalations, and you can send it back with a question. You can call a huddle, and the clock keeps running while you talk. The host sees the pressure on the room and what's coming up, and can watch any AI seat work: its desk as it sees it, its reasoning step by step, and a button to take the seat over. We rebuilt nine of these screens last week from what a live run showed people needed.</p><p><img alt="The Operation Center board for Riverside Flood, run 1, at hour 8 of 24: one bar per incident hour grouped by phase, what the scenario did on the left, and what the room is doing on the right, filtered to the four items that need you." src="assets/posts/simulate-video-and-the-loop/ops-board.png" loading="lazy"/></p><p><em>The board at hour eight. The scenario on the left, the room on the right, and the four things waiting on you.</em></p><p><img alt="The Resources Support Section Chief's desk: an incoming call saying all access roads are now impassable, tiles for Decide now, Escalated to you and For awareness, three escalations, and the reading pane showing the ask and why now." src="assets/posts/simulate-video-and-the-loop/ops-desk.png" loading="lazy"/></p><p><em>Your desk. A caller from the field on the line, three escalations, and the reading pane for the one you picked.</em></p><p>Running beside every room is a hollow run, a second simulation nobody acts on, so the debrief can set the world you got against the world you would have gotten by doing nothing. Findings come out in three lanes that are never added together: the people, the AI crews, and the plan. People's findings and the plan's can be filed to Improve. The crews' findings go to a benchmark and never become a gap in your record, because a machine's mistake is not your corrective action. There's an Experiments layer that runs the same incident with people and with crews under identical, locked and hashed conditions and measures the difference seat by seat and hour by hour, with outside reviewers and a kill switch, and you can register your own agent to play a seat. Any finished exercise can be run in an Operation Center with one click: its modules become operational periods, its injects go to the seats its stress map names, its discussion questions become decision points. Synthetic incidents only. Simulate Ops launches at IAEM26, and you can take a seat and try it there.</p><p><img alt="Watching the Mass Care Coordinator, an AI seat: the strip of every seat in the room, what the seat sees on its desk, and its reasoning step by step, with buttons to tell it something, pause it, or take over the seat." src="assets/posts/simulate-video-and-the-loop/ops-watch.png" loading="lazy"/></p><p><em>Watch a seat. What an AI crew sees, what it decided and why, and a button to take the seat yourself.</em></p><p><strong>Rehearse Decisions with SimCell</strong> is the just-in-time rehearsal I described two weeks ago. Describe a real situation, or start from a scenario starter, or let Preppr draft one from the news and your knowledge base. Name who responds, up to twelve seats, kept exactly as you wrote them. Give it an inject outline, typed or lifted from an exercise's MSEL. Check the plans before the run and it lists what they don't settle. Then it runs each decision you're weighing against a hold baseline and shows you the diff: where the plan parted from what happened, what the organization could not carry out when a plan asked the world for something it couldn't do, which tactics only worked on some paths. You can park every adversary move for review and accept, edit or withdraw it. You can continue from any path with the world state and the seats' memory carried forward. You can sweep one setting across its values and find the tipping point, with a verification re-run before the report is allowed to call it a cliff. It never recommends a decision. That sentence is in the help text and I intend to keep it there. What a rehearsal finds is filed to Improve as synthetic evidence that can corroborate a pattern but never create one.</p><p><img alt="The decision paths step of a rehearsal: each path runs as its own cell and is described as an option to compare, not a recommendation, with a hold baseline, seats that can voice concerns but never change the decision, and an option to review each Red inject before it lands." src="assets/posts/simulate-video-and-the-loop/rehearse-paths.png" loading="lazy"/></p><p><em>The last step before a rehearsal runs. The paths are options to compare, and the hold baseline is on by default.</em></p><p><strong>Adaptive Simulations</strong> now play graph-first with fog of war: the steps you've visited solid, the next hop as untitled options, everything beyond as silhouettes. Preppr is the game master. It maps your written or spoken response to a route and judges it pass, partial or fail with a rationale, citing the design brief, the exercise's sources, your knowledge base and the physics grounding stored at build, and a facilitator can confirm or override every verdict live. <strong>SimCell</strong> is where all of this started: Red, White and Blue cells in parallel runs that stress a plan before people do.</p><p>Four products, one principle. The model plays the world and the crews. The engine owns the truth. You keep the decision, and the clock makes you keep it on time.</p><h2 id="video-and-the-fence-i-was-on">Video, and the fence I was on</h2><p>Preppr can now generate video, and I was on the fence about it for most of a year.</p><p>Two reasons. The first was quality. I did not want to put a clip in front of a room of fire chiefs that looked like a dream sequence and made them laugh. An exercise lives or dies on whether the room takes it seriously, and one bad clip would cost more than a hundred good injects earned. The second reason was the opposite worry. If the quality got good, we'd be handing emergency managers a machine that makes convincing footage of disasters, and I have spent three years telling this field that the most dangerous thing about AI is believing it. Our own demo exercise has a presidential deepfake in it as an inject. I was not eager to build the thing that makes the inject.</p><p>The quality question answered itself this year. A short shot of a flooded intersection at night, with a narrator and the right ambient sound, now reads the way a news clip reads. That meant the second worry was the only one left, and the design followed from it.</p><p>Here is how the Exercise Media Studio works. Every scenario, module, inject and question in the designer has a Media button. A director agent reads the exercise and proposes a shot, a narration and a sound cue as priced cards with editable prompts. Nothing renders and nothing is charged until you click Generate on a card with its credit price on it. Corrections like make it night edit the clip you have rather than starting over. Every attempt is kept as a version. You pick one picture, one narration and one sound, set their volume and timing, and Preppr merges them into one clip with captions. Approve and attach, and delivery plays it to participants at that moment of the exercise. The same pipeline makes clips for the moments that matter in an Operation Center: the opening, decision points, phase changes and severe injects, each suggested by Preppr and ordered by the host, one at a time or all at once. Nothing renders unless a person asks for that clip. Media is billed at a premium rate you see before you generate, and it's on for selected accounts.</p><p>And every frame of it is labeled. Every generated clip and still plays under a persistent AI-generated, for training label. It is not a tooltip and it cannot be dismissed. Images Preppr generates carry a synthetic, exercise use only strip burned into the picture and the same disclosure written into the file's metadata, so even if someone crops the strip out, the file still says what it is. Anything that could pass for an official product, a mock news screenshot or a map, gets a diagonal watermark across the whole image on top of that. The rules every prompt is written under are not negotiable: training-appropriate only, no injuries, no real or identifiable people or officials, no real logos or news-network branding, no real addresses, preset voices only, and never anything designed to pass as real footage outside the exercise. Someone will ask whether the label can be turned off for a more immersive experience. No. The label is the feature.</p><h2 id="the-loop-you-can-see-and-the-page-that-closes-it">The loop you can see, and the page that closes it</h2><p>Every post on this blog eventually says the same thing: preparedness resets to zero after every exercise, and the Preparedness Cycle was always drawn but never ran. The two pages I'm proudest of this month are the ones that let you check whether that's still true of you.</p><p>The first is a card on the dashboard called your preparedness loop, which takes over from the workflow map once setup is done. Each run you finish closes a pass: a tabletop, a SimCell rehearsal, a Simulate Ops run, an adaptive simulation. Each pass is a row of cells, one for everything that fed it since the run before. Each document you added. Each question you asked Preppr. Each intel pull, each campaign, each design. Then the run itself as the deep tile at the end. Pass four, delivered October 21. Pass five, in progress since October 23. It uses the same visual grammar as the rest of the product now: bright green is what changed since last night, warm yellow is a gap or a contradiction, a ring is something a person set by hand, a dashed cell is a source Preppr hasn't read. Above it sit what Preppr has read, your readiness, one cell per capability that lights when a run tests it, and what Preppr noticed last night. There's one button, which opens a conversation already seeded with where you stand: since the chlorine release tabletop on the twenty-first you've added two documents, an Ask Preppr session and a new design, so design the next exercise around what they changed. Or, more pointedly: nothing new has come in since that run, add findings or documents, or schedule the next one. Nobody has to remember to run the loop. The loop tells you it's waiting.</p><p><img alt="The Your preparedness loop card: five passes, each a row of cells for the documents, Ask Preppr sessions, intelligence sessions and designs that fed it, ending in a darker cell for the run that closed it, with the fifth pass still in progress." src="assets/posts/simulate-video-and-the-loop/preparedness-loop.png" loading="lazy"/></p><p><em>Four passes closed and a fifth under way. Every cell is something that fed the run at the end of its row.</em></p><p>The second is Improve, which is where everything lands. Across the top, the numbers: exercises run, including the ones we only know about from after-action reports you uploaded, exercises designed, capabilities exercised, findings, priority gaps. Then capability coverage, one panel per framework you answer to, planned against tested, laid out as a map by mission area with one cell per capability. Hover a cell and it names the exercise that tested it. Click a grey one and it starts designing an exercise around it. A strip underneath shows how coverage has built up quarter by quarter, which is the first picture of compounding I've been able to put in front of a leadership team.</p><p><img alt="The top of Improve: four exercises run of five designed, eight of 32 FEMA core capabilities exercised, four open priority gaps, the capability coverage map with one cell per capability in mission-area rows, and how coverage has built up by quarter." src="assets/posts/simulate-video-and-the-loop/improve-coverage.png" loading="lazy"/></p><p><em>Improve, filtered to FEMA Core Capabilities. One cell per capability, and the quarter-by-quarter build-up underneath.</em> Then the priority gaps, each with its provenance: from an exercise, from an uploaded AAR, from a Collaborate campaign, from the people and plan lanes of an Operation Center, or from a rehearsal, marked as synthetic. Re-synthesize them, add your own, close one with Ask Preppr, or send the whole list off to develop an improvement plan. Beside that, the hazards in your intelligence feed you have never exercised, and your knowledge base's health, where an aging plan has a Review button that asks Preppr what's out of date and an after-action report has an Exercise button that turns its open corrective actions into the next exercise. And as of this week, you can just ask. Ask Preppr reads your improvement status and your preparedness metrics directly, so a dashboard of current improvement actions is a sentence, not a search.</p><p>I still say no to the AAR button. Improve is what we built instead of it. Not a document a machine wrote and nobody argued with, but a living list of what's been tested, what hasn't, and what's open, where every item says where it came from and hands off to the next run.</p><h2 id="giving-some-of-it-away">Giving some of it away</h2><p>Ask Preppr's Actions are now called Skills. Only the name changed: the THIRA, the CPG 101 plan evaluation, the after-action input analysis and the rest are right where they were. Two things around them are new. Your own AI assistant can now run them, whether that's Claude, ChatGPT, Copilot or anything else connected to Preppr through MCP, asking you the same questions Ask Preppr would and producing the same deliverable. And we've started giving them away.</p><p>There is now a free, public <a href="https://github.com/Preppr-ai/preppr-skills">Preppr Skills repository on GitHub</a>. The first skill in it is the one I've written about most: alert and warning template generation, built with researchers and risk communicators. It walks you through five alert types, Imminent Threat, Public Safety, AMBER, Missing Person and Developing Situation, each asking for what that type needs in the order the public expects to read it. Before anything is written, it checks your answers against plain-language, certainty and content-order standards, and anything bound for IPAWS or Wireless Emergency Alerts has to carry a clear protective action or it comes back to you to fix. Then it writes templates only for the channels you actually use: 90 and 360 characters for WEA, mass notification, email, voice and social. It installs as a native skill in Claude, GitHub Copilot, Gemini CLI and OpenAI Codex, or as a Gemini Gem or a ChatGPT custom GPT. It needs no Preppr account. It's licensed CC BY-SA 4.0, so use it, change it, and share what you change.</p><p>Why give it away? Both reasons are in the last post. Access is settled. Practitioners are already using general-purpose AI because it came with the license, and the real question is whether what they use is grounded in anything. If you're going to draft alerts in a general-purpose model anyway, I'd rather you did it inside a structure built from the research than from a blank prompt. And the measured improvement in alert quality has come from structure: templates written in advance, with experts reviewing, not live generation at incident tempo. A skill that builds the template before the incident is the version of this use case I think has earned its place. The problems in this field are bigger than any one company. This is a small way to act like it.</p><h2 id="the-fine-print">The fine print</h2><p>Simulate Ops launches at IAEM26, where you can try it in person. Rehearse, the Media Studio and the preparedness loop card are on for selected accounts while we watch them under real use. Operation Centers run synthetic incidents only. The Preppr Skills repository is free and public and needs no account. The screenshots in this post come from demo accounts with synthetic data. Generated media costs more than anything else in the product and says so before you spend. The <a href="https://changelog.preppr.ai">changelog</a> has the full account of what shipped, including the fixes, and it will keep having it.</p><p>The homepage says preparedness that compounds. This month is the first time I can point at a screen and show you yours doing it.</p>]]></content:encoded>
</item>
<item>
<title>The Fainting Goat Phase Is Over. Now Comes the Hard Part.</title>
<link>https://blog.preppr.ai/a/the-fainting-goat-phase-is-over/</link>
<guid isPermaLink="true">https://blog.preppr.ai/a/the-fainting-goat-phase-is-over/</guid>
<pubDate>Thu, 15 Oct 2026 12:00:00 GMT</pubDate>
<dc:creator>Justin Snair</dc:creator>
<category>Industry Thought</category>
<description>The state of AI in emergency management in 2026: from curiosity, to the fainting goat phase, to real pull, and the procurement, evaluation, and governance problems that come with it.</description>
<content:encoded><![CDATA[<p><em>The state of AI in emergency management in 2026: what I'm seeing, hearing, and learning. Notes from a few thousand conversations with practitioners, more than fifty presentations, and more contracts than I ever wanted to read. Nothing here is attributed. The people in these conversations didn't sign up to be quoted, so what follows is the pattern, not the person.</em></p><p>In 2023 and most of 2024, almost nobody in emergency management was using AI. But I was, and we were building. The question in every room was what is this thing. I'd ask who had used AI that day and about a fifth of the hands went up, and then I'd walk them through Netflix and Google Maps and the weather app until most of the room was in. The work that year was awareness: show people what was possible, build fast, let them poke at it. Curiosity, not adoption.</p><p>2025 was the year I started calling the fainting goat phase. You've seen the video. You walk toward the goat, it stiffens up, it tips over. That was every agency I approached. Funding got disrupted, national priorities were suddenly in question, and the field had no capacity to take on anything new. Say the words adopt AI and the whole organization locked up and fell over. They weren't being difficult. They needed some semblance of stability before they could consider anything, and the right move was to leave them alone and let them get back up.</p><p>2026 is different, and the difference is the subject of this piece. The appetite came back, and it came back as pull rather than push. This year I did the raise-your-hand test in a room of about 150 people and didn't need the walkthrough; two people said they hadn't used AI that day, and they were almost certainly wrong. The people calling me now aren't early adopters who'll tolerate something half-finished because they're curious. They have a problem, they've concluded that waiting it out isn't a plan, and they want to know how to buy. Which is where the new friction lives. The question in the room has moved from what is this thing, to how do I use it, to how do we procure it, and the third question turns out to be the hardest of the three. Procurement, contracting, security review, AI policy, and the vocabulary to evaluate any of it: that is the hard part, and it's where the field is stuck right now.</p><p>A note on where I'm standing. I run an AI company in this field, which means I have skin in the game and you should be skeptical of me. I'd rather you be skeptical and informed than trusting and burned. That's the choice the field is facing right now, and most of what I see every week is some version of it.</p><h2 id="1-what-the-modal-preparedness-office-looks-like">1. What the modal preparedness office looks like</h2><p>A county emergency manager opens a demo by saying the office is down to a one-person shop and nobody knows whether the second seat will be filled. Asked what's on the docket, the answer comes in order: get the grant work done, then the training the grant requires, then the actual work work.</p><p>That's the office I see most. One or two people. A multi-county coordinator covering three counties, three administrations, three emergency managers, with roughly half a person's time for preparedness. A state preparedness lead back after five years away, finding plans five and six years old and an exercise calendar already locked in for a five-year grant period she didn't write. A single office with seventy plans, most outdated, the first fifteen pages of each repeating the last.</p><p>The work plan is written by the grant, not the hazard. The grant deliverable will tell us what we're going to focus on is a sentence I hear in different words every month. The exercises get chosen years in advance to satisfy a funder, the after-action report arrives months after the exercise, and the people who understood the surge dependency have taken new jobs by the time anyone reads it.</p><p>Underneath that is a debt the field hasn't paid down. A lot of things were left undone during COVID and never picked back up. And 2026 is the year the federal floor moved. The council the President appointed to review FEMA delivered its final report in May and recommended a model it calls locally executed, state managed, federally supported, which in practice means pushing environmental review, procurement, eligibility decisions, and mitigation dollars down to the states, with most of the big changes waiting on Congress. The preparedness grant cycle itself went to court: new conditions attached to this year's preparedness and counterterrorism grants drew a lawsuit from more than two dozen states, a federal judge had already blocked an earlier round of grant policy changes, and as recently as the end of September a restraining order forced FEMA to pull back award notices and stop payments. Inside the agency, the government's own watchdog found that roughly a fifth of one class of disaster response employees wasn't renewed at the start of the year with no record of anyone assessing the effect, a hiring freeze ran until May, and a new requirement for cabinet-level sign-off slowed grant administration. State agencies responded the way you'd expect: vacancies held open, training and purchasing decisions made in a hurry or not at all. And the federal money that paid for hospital and coalition preparedness is being cut, which forces a choice nobody wants to say out loud: drop exercising and planning, or find a different way to do them.</p><p>I've said before that states and locals hold the ultimate responsibility here, and I still think that's right. But for a one-person office whose year is written by the grant, a grant that is simultaneously being restructured, conditioned, litigated, and delayed is not a budget. It's a weather forecast.</p><p>Emergency management is invisible unless something goes wrong. The public thinks we're in an ambulance. We're closer to insurance, and the budget shows it. So when a practitioner reaches for AI, it's rarely enthusiasm. It's we're underfunded, do this faster, better, cheaper. That's a real need. It's also not an answer to why the function isn't valued, and AI won't fix that part.</p><h2 id="2-the-first-question-is-always-the-same">2. The first question is always the same</h2><p>Before I've shown anything, the first question on most calls is how is this different from the Copilot we already have? That question is the tell. General-purpose AI is already in the building. It's on the government machine because it came with the license. It's our only approved AI, as one program lead put it, busy being angry at it because it couldn't build the pivot table she needed.</p><p>The pattern after that is consistent. Someone points the bundled assistant at the document repository, it looks good on the surface, and some weeks later they realize it isn't doing what they thought. It's pulling the old version. It's confidently summarizing a plan it only partly read. Then they go shopping for something purpose-built, usually on a personal device, because the network blocks the vendor's website.</p><p>Here is the mechanism, because practitioners deserve to understand it. Every model carries a bucket. The system instructions go in the bucket. Your documents go in the bucket. Every turn of the conversation goes in the bucket. The model runs laps between you and a server, and as the bucket fills, water spills out. It never tells you what fell out. Hand it a thousand-page hazard mitigation plan, which is about average, and it says I've got you while it has already dropped most of it. When the session gets long it offers to save a memory, which is a summary of a conversation it already lost the details of. Most people think the thousand-page plan is in there for good. It isn't.</p><p>And it won't say I don't know. By design, it's been rewarded in training to answer. So it flatters. Anyone who has used a consumer chatbot has been told they're the smartest person in the room. That's charming on an apple pie recipe and dangerous on an evacuation annex.</p><p>The lesson I've drawn: access is settled. The question is no longer whether practitioners will use AI. It's whether what they use is grounded in anything. Bans don't change the first fact; they only move the use to personal phones, along with whatever sensitive content goes with it. The policy that assumes use and governs it will beat the policy that pretends otherwise.</p><h2 id="3-the-real-game-isn-t-em-shaped-content">3. The real game isn't EM-shaped content</h2><p>Strip the marketing off most AI products in this space and you find a model pointed at an emergency management template. Plan-shaped output. Scenario-shaped output. Inject-shaped, AAR-shaped. It looks like the field. That's the surface game, it's mostly solved, and on its own it's close to worthless.</p><p>The underlying game is four things, and they're harder than the surface game by an order of magnitude. Complete context: the whole of what the organization knows, held without the bucket spilling, not whatever fit in the last session. Retrieval: getting the right thing back, which is a relevance problem before it's an accuracy problem; you can accurately retrieve something that has nothing to do with the question. Evaluation against a fixed corpus of rules and behaviors: doctrine, regulation, the organization's own commitments, and the patterns it has actually shown in exercise after exercise, so that an output can be checked against something that doesn't move. And surfacing when the language model is wrong or ungrounded. Ungrounded doesn't mean useless. In practice, the made-up thing might be exactly the counterfactual you needed. But you have to know it was made up, and you have to be the one who decides what to do with it.</p><p>Here's why that separation matters. Language models are good at language, framing, synthesis, and opinionated judgment. They are not good at keeping the truth straight. Ask one to draft a hurricane scenario for a hospital and it will sound exactly right: the cadence of incident command, the right acronyms in the right places. Then ask whether the population figure matches the county, which CMS regulations bind that facility, and what happens to fuel reserves at hour 72 of generator load. It will answer all three confidently. Two of those answers will be wrong, and you won't be able to tell which two.</p><p>So the model writes the story, and something that isn't a model has to own the truth and check the story against it. We built the company around that split: a deterministic engine of rules and hazard data underneath, language models on top, and an explicit flag whenever the story leaves the ground. Most vendors in this space didn't build that, and you can tell really fast. A chatbot writing stories feels one way. A system that handles context structurally feels another. The only similarity between us and some of the products I get compared to is that we both write words using AI.</p><p>The practical test for a buyer is one question. What happens when the model is wrong? If the answer is some version of it isn't, you've learned what you needed to know.</p><h2 id="4-where-the-time-actually-goes">4. Where the time actually goes</h2><p>A state champion finishes IT review through our public trust center and budgets three years at list price. The next email is from the contracting office, promising a year-long process. The three-year budget will now buy one year, a year from now.</p><p>Procurement is where AI hope goes to die. My life has become contracts and insurance. The timeline runs from weeks to eighteen months. I've watched a county take three months to land a four-figure subscription because its contracting office required its own commercial agreement. I've watched a large city spend three months approving a purchase that cost about as much as a streaming service. One state has been ready to buy for two years. Another killed a deal on a single sentence claiming the state owns everything the product generates, and afterward asked me what had killed it. One deal died after fifteen months because the risk manager quit, that signature was required for approval, and the position sat vacant. Fifteen months of work, undone by an org chart.</p><p>The clauses repeat. Uncapped liability and indemnification. Five years of record retention after termination. Multiple governments have required my AI to carry automobile insurance. Three-hundred-question security questionnaires. Terms of service that must be frozen, when software terms are living documents, so every frozen copy becomes a branch to maintain: one agency wants data deleted five days after cancellation, another is fine with a year. A request for proposals worth eight figures required six people in seats, which is to say a headcount contract that cannot buy software at all. Nobody has figured out how to bill for an agent in six-minute increments.</p><p>None of this is malice. Government contract templates were written for labor, not software, and long before anything called AI. They're written for the buyer, not bilaterally, and that made sense when the vendor was a staffing firm. Emergency managers don't come from legal or procurement, yet contracts, mutual aid agreements, and vendor terms are what actually get you the stuff in a disaster. That's a gap worth exercising, by the way.</p><p>Some encouraging things. Security review is now universal and nearly identical from the smallest county to the largest enterprise: models, data flows, retention, hosting. Transparency clears it. A public trust center with requestable architecture documents does more than any sales deck. A few states have built formal AI safety reviews, and passing one is becoming a portable credential: one state's committee reviewed us unprompted for four months and then authorized any state agency to adopt without further AI review. That's governance working. Meanwhile other states have a review requirement and no submission process. You have to find someone to do it.</p><p>What I've concluded: the binding constraint on AI in emergency management in 2026 is not model capability and not practitioner appetite. It's contract language. Roughly 70,000 jurisdictions each have their own secret thing, and the vehicles are being rewritten mid-deal. The fix is boring: standard clauses for AI software, state-level review gates with a front door, and vendors willing to publish enough that review has something to review.</p><h2 id="5-what-people-ask-me-about-accuracy-and-what-i-tell-them">5. What people ask me about accuracy, and what I tell them</h2><p>On a demo, a prospect asks what's its accuracy and waits for one number. I say around ninety percent, depending on what you're asking it to do. Wait, it's not just one number? And I watch some of the trust drain out of the room while I explain the trick.</p><p>Accuracy isn't a number. It's like saying Justin is accurate. It moves with the task, the corpus, the question, and how tired the model's context is. A system can be ninety-five percent accurate at retrieving the right section of a plan and seventy percent accurate at summarizing what the section says. Borrow the intelligence community's language instead: confidence, never truth. This is thirty percent likely to be true. Make your choice.</p><p>The ceiling is real. No lab, with billions of dollars, has pushed a language model past the low nineties on the benchmarks it publishes, and hallucination is a mathematical property of the tool, not an engineering bug. The best I can do is say when the system is making something up. So when I mean ninety-three percent on a specific measurement, I say ninety-three percent. When a vendor leaves the percentage out, they are telling you it's a hundred. Trust me, bro.</p><p>A 95 percent accurate AI in marketing is survivable. A 95 percent accurate AI in public health preparedness or emergency management carries 5 percent of possible harm, and that 5 percent is not survivable. But it's not one bar. A 90 percent accurate AI drafting an after-action report is a productivity win; errors get caught in review. The same model generating a wireless emergency alert during an incident is a liability event waiting to happen. One alert in ten wrong, pushed to hundreds of thousands of phones, with no recall. Emergency managers already think this way about every other piece of equipment: a HazMat detector that's 90 percent accurate is fine for perimeter screening and dangerous for clearance. AI is no different. The accuracy bar has to match the consequence of the decision it's informing, which means the field doesn't need a benchmark. It needs a benchmark family.</p><p>Alerts are where this argument gets tested, so let me take the strongest version of the other side. Some will say AI-generated alerts only need to be better than the typically bad ones humans write, and on the surface that's true. I've read the typical human alert: acronyms nobody outside the agency knows, no clear protective action, a reading level that loses half the audience. The research backs that up. A review of 6,080 wireless emergency alerts sent between 2012 and 2022 found that only 8.5 percent included all five things a warning needs: the hazard, the location, the source, protective action guidance, and the time. Source and timing were the most often missing (<a href="https://doi.org/10.1111/1468-5973.12518">Olson, Sutton, Cain, and Waugh, 2024</a>). A model fixes most of that on the first pass.</p><p>The comparison hides three things. First, it compares rates, and harm is a count. Harm is volume times error rate times what each error costs. AI lowers the cost of producing an alert, so volume goes up: more platforms, more languages, more updates, finer zones. Say a human writes twenty messages across an incident and gets six wrong, and a novice with a generator writes two hundred and gets twenty wrong. The rate improved and the N of harm went up. Those numbers are hypothetical, because nobody has published a real one. Second, the errors stop being independent. A human's mistake is one message on one channel. A generator's mistake is one wrong zone pushed to every platform in both languages at once. The error is cheaper and it travels further, and the human review that would catch it doesn't scale with the volume, so the effective error rate drifts toward the raw model rate as output rises. Third, better than humans at what? Human alerts fail mostly on completeness, and that failure is well measured. Models fail on facts: wrong zone, wrong time, wrong contact, a website address with an extra letter in it, and nobody has published how often that happens, for humans or machines. The improvement is on the axis we've measured. The new risk is on the axis we haven't. A single comparison hides the trade. None of that counts alert fatigue, where volume itself erodes the public's response to the next message, correct or not.</p><p>I'm not saying don't use AI for alerts. We built an evidence-based alert template tool with researchers and risk communicators and gave it away. I'm saying evaluate that use case on its own terms. That means a benchmark for alerting, stratified by failure mode, that measures total harm across an incident rather than a per-message rate, against a human baseline that someone has actually measured, because right now nobody has measured factual error rates on either side. And it means friction between generation and transmission that grows with the volume: a hold, an approval, a named human. It's also worth noticing where the measured improvement in alert quality has come from: structure. In controlled experiments, messages written to a standardized template outperformed typical messages on understanding, belief, and confidence about what to do, before any model was involved (<a href="https://doi.org/10.1061/NHREFO.NHENG-2293">Sutton and colleagues, 2025</a>). Pre-scripted templates built in advance with experts reviewing are one thing. Live generation at incident tempo is another, and the second hasn't earned its place yet.</p><p>Any vendor can claim accuracy. Any vendor can claim safety. Any vendor can make up great-sounding terms, like closed circuit AI, that mean little and differentiate less. Drill into closed circuit and it usually means not connected to the internet, while the data still flows to a model provider. I use fourteen models and I say so. I've read terms of service in this space that still permit training on customer data, and I've watched a competitor change its terms in the middle of a trial without showing what changed. I've watched an emergency operations assistant agree to role-play a pirate and invent answers within seconds. Vendors who sell safe, accurate, closed circuit with no qualifier aren't earning trust. They're renting it, usually from a relationship.</p><p>There's a trap here for anyone honest, and I've fallen into it. The more I explain where the AI is wrong, the less some buyers trust me. There is research on this: the person who explains the magic trick loses credibility even when he built the trick. More than once I've thought I should have just shut up. I'm simultaneously selling AI to these agencies and telling them the most dangerous thing they're about to do is believe it's perfect. I don't want to rent their trust. I want an informed purchaser. That costs me deals, and I'd make the same call again.</p><h2 id="6-the-pull-toward-response">6. The pull toward response</h2><p>Mid-demo, a hospital preparedness coordinator cuts in: "You keep saying practice, practice, practice. From what I see, this is very accurate on our real data. Why couldn't this run during a real activation?" I tell her it's a philosophical line. She shrugs. All she wants is for the system to whisper that could be wrong when someone on the call says something the record contradicts.</p><p>The pull toward response is universal. Partners tell me there is infinitely more interest in anything related to live response than to practice. Search interest in testing is falling while training rises, and customers ask for just-in-time reps. Investors and advisors tell me I have a response platform and don't know it. Academics ask whether the data will include real events or only exercises. A response demo is thrilling. A preparedness demo is a person explaining their planning process. The money and the visibility are on the response side, which is exactly the side where a probabilistic tool is least appropriate.</p><p>My position hasn't changed, but I want to state it precisely, because the imprecise version gets quoted back to me. The design goal is to create the most space possible between a language model generating something and a human being harmed. Preparedness gives you weeks, months, or years to catch the wrong answer. Response gives you minutes. A language model under duress is all possible answers to all possible questions at all possible times, and the moment a person most needs to scrutinize the output is the moment they're least likely to. I think that's a human problem more than an AI problem: under pressure, people franchise out their decision-making. I won't build a tool that loads a gun and points it at an emergency manager.</p><p>So: AI hasn't earned response yet. The word yet is doing honest work in that sentence. There is real value in information work around a response, synthesis and drafting and tracking and retrieval, and that is different from a tool near a time-compressed decision; we should stop letting the first launder credibility for the second. Guided help, where the playbook is and who to call, is different from a recommendation. Just-in-time rehearsal that lets a team test the difference between two decisions is different from telling them which to make. Business continuity, where nobody is dying, is a different consequence space than an evacuation. Deterministic rules engines and physics models, which have math behind them, are a different class of tool than a chatbot. If an independent evaluation ever shows a system performing better than the human alternative on a specific response task, with the mechanisms to prove it, entering that task would still be principled. Until then, the line holds, and I'm against my own product's expansion in that direction more often than my advisors would like.</p><h2 id="7-the-aar-button">7. The AAR button</h2><p>Users ask for it on nearly every call. One after-action report, please, from the exercise we just ran. I keep saying no, and I owe people a better explanation than principle.</p><p>Three reasons. First, the easy button is the problem. An AI-written report lets an organization certify that learning happened without any learning happening, faster than before. Second, a vendor-written final report is discoverable. When something goes wrong after a disaster, I don't want the authoritative record of your readiness to be a document a machine wrote and nobody argued with. Third, the public corpus of after-action reports is poor evidence. A national review of thousands of AAR-shaped documents found almost none usable as evidence of anything; peer-reviewed work has found most of them generic and many missing whole capabilities. Public AARs contain the three things that were safe to say. Point an AI at that library and you get the safe findings, faster.</p><p>What we do instead is the verb. The system captures the exercise at full fidelity: who said what, in which breakout, against which objective. It evaluates by conversation, with the facilitator defining what good looks like and approving or rejecting every candidate finding. And it does one thing I've come to think is the most valuable feature in the product: it compares the recorded exercise to the sanitized report that followed it and asks, why does this authoritative document say nothing like what was said in the room? We've sanitized so much along the way that it's fair to ask what we're improving from. The data that matters is the conversation that never made it into the document.</p><p>Two more things I've learned from exercise people this year. A veteran designer told me clients say they hate the evaluation paperwork, and what they mean is they don't want to plan the exercise carefully against a risk assessment and their plans. He compared it to sandpapering before you paint. And a learning specialist designing a competency assessment kept repeating that situational awareness and decision rationale are not quantitative unless you assign a scale and an evaluator who knows why the participant decided what they decided. Both are right. A tabletop tests a plan. It doesn't test time to decision or individual competency, and most cyber tabletops I see are drills. Europe is well ahead of us on actual simulation. The United States is still write a document and talk at each other.</p><h2 id="8-what-s-working">8. What's working</h2><p>A state preparedness lead listens to a demo and reframes it for me: so you're not an AI company. Right. I'm a context-engineering company with AI on top. The AI is not the problem anymore. The problem is where all the information is, and it isn't where we need it.</p><p>Here's what I see working, across the small number of organizations that are getting this right.</p><p>They define the problem first. We want to get into AI isn't a problem. We have seventy plans and can't tell which assumptions still hold is. When the problem is named, the tool follows. When it isn't, I tell people to use a spreadsheet. I now decline most inbound demos, because the organization either hasn't recognized a problem or isn't allowed to use AI, and a demo was always a vanity metric.</p><p>They treat plans as documented assumptions. A plan is a set of claims about reality frozen at the moment of signature, and reality drifts. The useful work is finding the delta between what the doctrine says, what the plan says, and what the organization actually does. Doctrine says seventy-two hours of fuel. Three exercises in a row say a hundred and twenty. I want the delta, and I want it surfaced continuously, not rediscovered annually.</p><p>They gather context at scale before they design. You could never get a thousand people into a planning meeting. You can interview them with an agent, collect their documents, and look for dependencies, consensus, and contradictions before the first draft exists. Then run a half-hour staff-meeting exercise against the gaps, and let the discussion drive the plan revision. Analysis, then practice, then findings, then revision. The old order was committee, then document, then shelf.</p><p>They understand the bar depends on the stake. The same ninety percent that's a productivity win on a draft is a liability event on a public alert. They ask for a confidence, not a verdict.</p><p>They push the button. The first user in every new account gets a note from me that says, more or less, you are not operating a nuclear plant, load a document, you'll be fine. People who have survived a full procurement cycle then stall at the first screen, afraid to break something. Reluctance to be wrong inside software is universal, but it runs deeper in a profession where being wrong has consequences. Adoption isn't a procurement problem. It's a first-click problem.</p><p>And a few quiet signals. A financial-services enterprise cleared our product faster than most governments. States are competitive in a specific way: nobody wants to be first, nobody wants to be last, everyone wants to be second, which means one honest public write-up moves ten agencies. And honesty about limits is itself a differentiator, because almost nobody else is offering it.</p><h2 id="9-the-market-and-the-room">9. The market, and the room</h2><p>Two hundred people, senior leaders on stage, and the words trust, human in the loop, and performance standards repeated all day without anyone defining them. That was my experience at the AIDE Summit this year, and I want to be careful about how I say this. The people behind that effort are doing essential work. They mapped the market when nobody else had, they are leading the charge for responsible adoption, and the field is better for it. I think they could have gone further, is all. I had hoped for much deeper dives, and instead much of the conversation circled the same undefined notions that get repeated in every room I'm in. By the afternoon I was frustrated.</p><p>So when I got on stage, I decided to speak as honestly as I could, and to accept that I may never be invited to speak again.</p><p>The best count we have says there are about 1,179 AI-enabled products for emergency management from 717 companies. Most are general-purpose tools pointed at the field. One in five depends on another product to function. Coverage is heaviest in response and thinnest in preparedness, which mirrors where the funding has always gone. And multi-organization coordination, the defining feature of how emergency management actually operates, is the thing the market has built the least support for. Lock-in is a better business model than integration. A thousand rational commercial decisions made the field harder to coordinate.</p><p>2026 is also the year the field got crowded. More entrants are showing up in AI software for emergency management every month, and I think that's great. I advise some of them, and I encourage others to get in, because the problems here are bigger than any one company. The risk that comes with a crowd is slop: thin products with confident marketing that pull attention and budget away from the work that matters, and make it harder for a buyer to tell the difference. Which is why we need evaluation standards now more than ever, so customers can compare tools fairly instead of on charisma. I'll admit the other side of it too. Slop is a great differentiator. When most of what's on the market is a wrapper and a slogan, a product that shows its work doesn't have to shout.</p><p>I'd rather say this before than after: parts of this field are going to get burned by what they're adopting right now. Not because the technology is all bad (though some really is), and not because anyone acted in bad faith. Because we are making high-consequence purchases on charisma, and some fraction of those decisions will fail in public at the worst possible moment. When that happens, the damage won't stay with the jurisdiction that bought the tool. It will land on the whole idea of AI in emergency management and set the careful work back by years.</p><p>We vendors were asked to agree on performance standards so agencies could compare products. I said what I think: you've pointed at five competing companies to set standards for processes the field hasn't proven in twenty years. We don't have a shared vocabulary for accuracy, relevance, or trust. Accuracy in a process that hasn't been shown to achieve anything doesn't matter. If what we're building is for increasing the production of paper, we're useless. AI adoption is not the problem. What are you trying to solve?</p><p>Two other things I'm watching. Practitioners can now build software without an engineering team, and some are shipping emergency management tools that way. You can build really fast and go to market. The question is whether you should. We're vibing our way to more theater, and the field sees AI as one thing, so one headline about one bad tool will undo years of trust-building for everyone. I've spent three years getting this field comfortable with AI, and I'm aware that the next setback won't care whose product caused it.</p><h2 id="10-governance-is-being-written-in-real-time-without-us">10. Governance is being written in real time, without us</h2><p>A national disaster-relief organization sends over its new AI policy as a condition of a free pilot. It reads like a prohibition. It was written without talking to anyone who builds this technology or anyone who uses it in the field. After a redline, they scrap it and start over. I've seen the same document, more or less, at a large technology company and at several agencies.</p><p>There is no AI governance in emergency management comparable to what healthcare has. Most existing policies are poor, and the honest version of the problem is that we don't yet know what we're governing. IT's incentive is to take on no risk it can't mitigate, and the easiest way to do that is to say no. So staff read the policy as a ban, keep using the tool on their phones, and a procurement question I now ask early is simply whether the organization has a policy that prohibits AI, because the answer decides whether we're having a real conversation.</p><p>Liability is the issue underneath, and it's the one nobody wants to know about. A veteran editor in this field told me it's the biggest problem precisely because nobody wants to find out the answer. A former government communications official described the scenario to me plainly: one catastrophic event, an emergency manager who says I used this platform, and a legislative hearing that subpoenas every chat log. Lawyers inside the government tell me the laws aren't adaptable enough for the technology. Everyone is waiting for the first case.</p><p>Meanwhile a federal team that quietly dropped a conversation with me two years ago, because they weren't allowed to use AI products, came back this year to say the restrictions have been lifted by a new memo. Policy moves in jumps. The emergency management function is almost never in the room when those jumps get written, and it is the function that will inherit the consequences. Participation, not abstinence.</p><h2 id="11-what-i-m-learning-about-my-own-position">11. What I'm learning about my own position</h2><p>I wear a shirt that says it's the AI's fault. It's intentionally ironic, because you can never say the AI did it. Accountability can't be delegated to a tool.</p><p>The thing I keep relearning is that this is harder than selling technology. I understand selling to government; it's the AI part the field isn't ready for. Showing most emergency managers an agentic workflow is like showing a caveman a lighter. The field is about two years behind what's possible, and most of it still thinks a language model is a word generator or a fancy search engine. That's not an insult. It's the state of the evidence and the state of the training. A workforce survey of what practitioners actually know, and what they do with these tools, is only now being drafted. We're building the field's capacity to adopt AI into a discipline that hasn't finished defining itself.</p><p>I've also learned to say no more. We don't build process when it isn't earned. I don't build a feature because a prospect's logo wants it. I've declined to put the product in rooms that weren't ready, and I've declined revenue tied to headcount, because the whole point is a change of labor, not a faster typewriter.</p><p>And I've learned that the sequence matters more than the tool. Data first: get the organization's own records out of the filing cabinet and the shared drive and into a form a machine can hold without losing it. Decision support second. Anything that deserves the word intelligence third. The organizations that get the sequence wrong buy AI they can't use, and the output hides the gaps instead of revealing them.</p><h2 id="12-what-i-think-it-adds-up-to">12. What I think it adds up to</h2><ul><li>The modal preparedness office is one or two people whose year is written by the grant, running on pre-COVID plans, with after-action reports that arrive months late and are read by nobody.</li><li>Shadow use is the default. General-purpose assistants are everywhere because they came with the license, and bans move the use to personal devices.</li><li>The binding constraint on adoption in 2026 is contract language and procurement, not model capability or practitioner appetite.</li><li>Security review has become universal and uniform. Transparency clears it. Secrecy doesn't.</li><li>The field has no shared vocabulary for accuracy, relevance, or trust, so unqualified marketing words are heard as a hundred percent.</li><li>The real game isn't EM-shaped content generation. It's complete context, retrieval, evaluation against a fixed corpus of rules and behaviors, and surfacing when the model is wrong. Most vendors are playing the surface game.</li><li>Silent context loss on long documents is the most common failure mode, and almost nobody recognizes it when it happens.</li><li>The pull toward response comes from money and visibility, not from evidence that it's safe there.</li><li>The after-action corpus is sanitized receipts. AI pointed at it reproduces the safe findings faster. The valuable data is the conversation that never made it into the document.</li><li>Plans should be treated as documented assumptions, and the job is finding the delta continuously.</li><li>Scale is the new harm vector: a ninety-percent-reliable generator in a novice's hands multiplies errors, not just output.</li><li>Honesty about limits is a differentiator because almost nobody offers it, and it costs the vendor credibility in the demo.</li><li>Governance is being written state by state, sometimes with no front door, and the emergency management function is rarely in the room.</li><li>The evidence gap is double. No one has proven the underlying practices produce outcomes, and no one has measured what the workforce does with AI.</li><li>The field treats AI as a monolith, so one badly built tool is a shared risk for everyone.</li></ul><h2 id="the-bottom-line">The bottom line</h2><p>Start with the problem. See if technology helps. Check whether you have the governance to do it safely. Find out whether the tool actually needs AI in it. Then decide if the risk is worth it. AI is not magic; it's a utility with a bill that scales with use. If pen and paper fixes it, you don't need it. If you don't understand the problem, stop.</p><p>I sell AI, and I'm telling you not to buy it unless it's necessary to do the work. In a field with no standard, where the cost of being wrong is measured in people, that's the only honest position available. Not everyone around you is going through this process, and eventually, even if you opt out entirely, you'll be an island surrounded by others using it.</p><p>The question isn't whether AI arrives in emergency management. It already has. The question is whether we'll know what we handed it, and whether it worked. Better compared to what? Let's go find out.</p>]]></content:encoded>
</item>
<item>
<title>Before We Ask Whether AI Works for Emergency Management, We Have to Know What Works</title>
<link>https://blog.preppr.ai/a/the-answer-key-jem-ai-series/</link>
<guid isPermaLink="true">https://blog.preppr.ai/a/the-answer-key-jem-ai-series/</guid>
<pubDate>Wed, 09 Sep 2026 12:00:00 GMT</pubDate>
<dc:creator>Justin Snair</dc:creator>
<category>Industry Thought</category>
<description>The Journal of Emergency Management is building a three-volume body of knowledge on AI in emergency management. The hardest question it has to answer isn't about AI. It's about us.</description>
<content:encoded><![CDATA[<p>I'm writing this from Austin at the AIDE Summit. And on the same morning, the Journal of Emergency Management announced something I've been arguing for in public for years: an independent, rigorous, peer-reviewed knowledge base for AI in EM. It's a three-volume scholarly and professional reference series, <em>Artificial Intelligence in Emergency Management and the Disaster Sciences</em>, and I think it's exactly what the field needs.</p><p>I've signed on as the Senior Technical Editor, so I'm not a bystander here. But the reason I said yes has less to do with AI than most people would expect.</p><h2 id="what-was-announced">What was announced</h2><p>JEM, led by Managing Editor and Publisher Richard A. DeVito, Jr., is developing the series around a simple progression: UNDERSTAND, APPLY, ADVANCE.</p><ul><li><strong>Understand</strong> builds the foundations: AI concepts, human-AI collaboration, ethics, bias, transparency, privacy, cybersecurity, governance, organizational readiness, and workforce competencies. The common language the profession doesn't have yet.</li><li><strong>Apply</strong> works through the mission: EOCs, situational awareness, forecasting, logistics, public warning, healthcare, public health, EMS, fire, critical infrastructure, damage assessment, training, and community resilience, across mitigation, preparedness, response, and recovery.</li><li><strong>Advance</strong> is the part I care most about: evidence, validation, professional standards, emerging technologies, research priorities, and the unanswered questions.</li></ul><p>The editorial team includes Attila J. Hertelendy, PhD; Sonny S. Patel, PhD, MPH, MPhil; Megan Snair, MPH; Emma Erwin, MPH, MSc; and Brenda DeVito, PTA. JEM is now inviting scholars, practitioners, researchers, educators, and subject-matter experts to participate as chapter authors, reviewers, and research collaborators.</p><p>That's the announcement. Here's the problem it has to solve.</p><h2 id="we-don-t-know-what-to-hand-over">We don't know what to hand over</h2><p>Every conversation about AI in emergency management eventually arrives at the same question: what should AI do, and what should stay human?</p><p>It sounds like a question about AI. It isn't. It's a question about us. To answer it, you need to know which parts of the practice are understood well enough to be specified, measured, and checked. Most of emergency management never has been. The profession runs on doctrine, tradition, and professional judgment, most of it never independently tested against outcomes. That's not a knock on the people. It's the state of the evidence. We can't tell AI what to do reliably because we've never verified what works. And you shouldn't automate a practice you can't evaluate.</p><blockquote>You shouldn't automate a practice you can't evaluate.</blockquote><p>Take exercises. HSEEP has shaped how this field practices for more than twenty years, and in all that time it has never been independently verified as effective. Not once. I think it works. Prove it. Planning is the same story. CPG 101 tells you how to build a plan, step by step, and the field follows it. It can't tell you whether plans built that way hold up when the disaster arrives. No one has checked that either. The same holds for how we assess risk, how we write after-action reports, and how we decide a jurisdiction is ready. We have process frameworks. We have very few verified outcomes.</p><p>So when an AI tool speeds up any of that work, we face a question we can't answer: did it make the work better, or just faster? Better compared to what? And the question under that one: are we making something faster that shouldn't exist at all?</p><h2 id="i-know-because-i-did-it">I know, because I did it</h2><p>Preppr started as an exercise design platform. I understood exercise design deeply. HSEEP existed as a process. I could describe it, specify it, and encode that expertise into the product. That's what made exercise design automatable in the first place: a practice well enough understood to be written down. Specified, in other words. Not verified.</p><p>But look at who wrote it down. What the AI does at scale, across thousands of exercises, was largely defined by me. It had to be. For AI to do a task repeatedly well, someone has to decide what "well" means, and there was no verified standard to hand it. So I made the calls.</p><p>And even now, there is no evidence to say whether a one-hour exercise run eight times is better than one eight-hour exercise. Not better according to post-exercise vibes from participants. Better against the highest bar: outcomes when the disaster arrives.</p><p>That's the gap. Not whether AI can do the work. Whether anyone has proven the work is the right work.</p><p>Exercises are just the corner of the field I know best. Now ask the same question everywhere else. RAND mapped 45 emergency management task areas in the AIDE report. Practitioners named 40 use cases for AI. Go down that list, planning, threat and hazard assessment, situational awareness, public warning, resource management, damage assessment, response, recovery, and ask honestly: have we proven the best approach to any of it? Not vibes. Not feelings. Not "I've done this for twenty years and I just know."</p><p>For most of that list, the answer is no. Not because the approaches are wrong. Because no one has checked. Before we add AI to any of it, the field has to be willing to ask that question out loud, and to be uncomfortable with the answer.</p><h2 id="what-ai-can-take-and-what-it-can-t">What AI can take, and what it can't</h2><p>The honest map looks something like this.</p><p>Where the practice is well specified and the output can be checked against something we already trust, AI is ready to do real work: synthesis, drafting, translation, retrieval, cross-referencing plans against requirements, tracking who has practiced what. This is administrative and analytical load, and it is exactly where AIDE's practitioner workshops found the strongest demand. The rule is clear, the output is checkable, the human stays accountable.</p><p>Where the practice is unverified, the risk changes shape. Adding AI to unproven practice is a risk of its own. You don't get better. You get faster. Faster plans, faster exercises, faster paperwork, and no idea whether any of it moved an outcome in a disaster. Speeding up an unproven process doesn't improve it. It scales it. An AI that produces ten times the exercises against a design methodology no one has validated produces ten times the unknowns. The bar can't be throughput. It has to be outcomes.</p><blockquote>You don't get better. You get faster.</blockquote><p>And where the work is judgment under uncertainty, with lives downstream of the decision, AI hasn't earned the seat. Not because the models can't produce an answer. Because we can't yet say what a good answer is, or prove one was made.</p><p>The line between these categories is the most important thing this field needs to draw, and right now it's being drawn by how we've always done it, procurement language, and product demos. It should be drawn by evidence.</p><h2 id="why-a-journal-and-why-now">Why a journal, and why now</h2><p>The AIDE report says the thing holding adoption back isn't a shortage of tools. It's trust. That's right, and it raises a harder question: how do you scale trust? One jurisdiction can trust a tool because they piloted it, or because they know the people who built it. A whole field can't. Trust at that scale has to rest on something everyone can inspect: reliable, verified, evidence-based approaches, and AI held to the same standard. Trustworthy because it was tested, not because someone said so.</p><p>Drawing that line is research. It needs practitioners who know the work, academics who know how to test it, and technologists who know what the tools can actually do, working together and publishing what they find where it can be cited, reviewed, and argued with. It has to be built independently, and it has to be contestable. A body of knowledge earns authority by being open to argument. That's the only thing that turns "we think this works" into "we know what works, and here's what we still don't."</p><blockquote>A body of knowledge earns authority by being open to argument.</blockquote><p>The timing isn't optional. Policy, funding, and education are already behind the innovation curve. Tools ship faster than guidance, and guidance ships faster than evidence. That gap widens on its own. The only way it closes is if the field deliberately builds the evidence base, and does it now, while adoption is still early enough to be shaped by it.</p><h2 id="what-i-m-asking">What I'm asking</h2><p>If you're a scholar, a practitioner, a researcher, or an educator with real expertise in AI, emergency management, or the disaster sciences, JEM wants you: chapter authors, reviewers, research collaborators. <a href="https://wmpllc.org/ojs/index.php/jem/ai-in-em-dr-ds">Details and how to participate are here.</a> The work is slower than a product launch and harder than a conference panel. It also outlasts both.</p><p>The question isn't whether AI will enter emergency management. It already has. The question is whether we'll know what we handed it, and whether it worked.</p><p>Better compared to what? Let's go find out.</p>]]></content:encoded>
</item>
<item>
<title>1,179 Products and No Agreed Upon Dictionary</title>
<link>https://blog.preppr.ai/a/1179-products-and-no-agreed-upon-dictionary/</link>
<guid isPermaLink="true">https://blog.preppr.ai/a/1179-products-and-no-agreed-upon-dictionary/</guid>
<pubDate>Thu, 03 Sep 2026 12:00:00 GMT</pubDate>
<dc:creator>Justin Snair</dc:creator>
<category>Industry Thought</category>
<description>What the AIDE report gets right, and the question underneath it that predates AI entirely.</description>
<content:encoded><![CDATA[<p>It's National Preparedness Month, and next week I'll be in Austin for the AIDE Summit, speaking on a panel about AI approaches purpose-built for emergency management — alongside Sean Griffin, Tony Fung, Joel Wish, and Tyler Felous, moderated by Matthew Watkins of the Texas Tribune.</p><p>I'm glad it's that group. Every one of us has had to make these calls with our own roadmap and our own money: what to build, what to refuse to build, which part of the lifecycle to serve, and what to tell a customer who wants something the technology can't responsibly do yet. Those are concrete decisions with consequences attached, and they make for a more honest conversation than a panel of people theorizing about the market from outside it. Builders should be held to a higher standard of specificity here, not a lower one.</p><p>The AIDE report — <em>AI for Disasters + Emergencies: A Way Forward</em>, from the Markle Foundation with Aspen Digital and RAND — came out at the beginning of August. Normally I'd have posted my reaction to something like this within a day or two. I've been sitting with this one for a month instead.</p><p>It earned that. 1,179 AI-enabled products from 717 companies, mapped against 45 emergency management task areas, alongside months of practitioner workshops. It is the most serious assessment we have of where this field actually stands, and a fast take would have been the wrong response to it. Some documents are worth reading twice before you say anything. If you work anywhere near this intersection, read it.</p><p>What follows is what a month of sitting with it produced. I want to do two things: credit what it gets right, which is a lot, and then sit with a question it doesn't quite ask — one that has nothing to do with AI, and everything to do with whether we'll be able to tell if AI is helping.</p><h2 id="what-the-report-gets-right">What the report gets right</h2><p>Two findings have stayed with me.</p><p><strong>The market is thinnest in preparedness.</strong> The report says plainly that this mirrors where our funding has always gone. Product coverage is heaviest in response, moderate in recovery, and thinnest in the work that happens before anything goes wrong. We have under-resourced preparedness for decades, and the product landscape simply hands that back to us.</p><p>And this line, which I'd put on a wall: <em>"Multi-organization coordination is weakly automated across the product landscape, despite being the defining feature of how emergency management actually operates."</em></p><p>That is the sharpest sentence in the document. The thing our profession is for — getting organizations that don't report to each other to act coherently under pressure — is the thing the market has built least support for.</p><p>The report's diagnosis is that useful tools exist but agencies lack the capacity and trust to use them, and its answer is a capacity-building agenda: a summit, an education program, a playbook, capacity-building pilots, a research agenda, and an alliance. I support all of it. The education piece is where I'd invest hardest.</p><h2 id="why-i-built-in-preparedness">Why I built in preparedness</h2><p>The thinness finding is why I built where I built.</p><p>There was ample encouragement to go sell into response. That's where the attention is, where budgets move fastest, and where demos look best. A response demo is thrilling. A preparedness demo is a person explaining their planning process.</p><p>I don't think AI has earned response yet.</p><p>Response is high-consequence, time-compressed, and unforgiving of confident wrong answers. We are nowhere near the evaluation standards that would justify putting a model near those decisions. There is real, working value in AI supporting the information work around response — synthesis, drafting, tracking, translation. That is different from AI near a time-compressed response decision, and we should stop letting the first one launder credibility for the second.</p><p>Preparedness, by contrast, is unglamorous, chronically under-resourced, and genuinely improvable right now. So that's where I went, and I'd make the same call again.</p><h2 id="the-coordination-gap-has-a-cause-the-report-doesn-t-name">The coordination gap has a cause the report doesn't name</h2><p>Coordinating across organizations requires software that talks to other software. For two decades, the commercial incentive in this space has run the other way.</p><p>Lock-in is a better business model than integration. That is the whole explanation. It's how you get a landscape where every tool owns a slice and almost none hand off cleanly — and where RAND found that one in five products depends on another product just to function.</p><p>The report describes the coordination gap as a market fact. I'd name it as a market design. Nobody set out to make the field harder to coordinate; a thousand rational commercial decisions did it anyway.</p><p>We went the other direction — open interfaces, APIs, agent-facing standards like MCP — so a customer's data and workflow can leave if they want them to. We're not the only ones anymore, and that shift is the most encouraging thing I've seen in this market in years.</p><h2 id="the-question-underneath">The question underneath</h2><p>Here's what I keep circling, and it has nothing to do with AI.</p><p>Preparedness produces documents that get marked complete. Plans, exercises, training records, capability assessments. Almost none of it is systematically evaluated against agreed standards. Mostly we count that something happened. An exercise was held. A plan was updated. A capability was worked on, a capacity was discussed. Very rarely is any of it tied back to how the response actually went.</p><p>So when the report asks — as its research agenda properly does — whether AI improves the work, I'd ask the prior question:</p><p><strong>Improves it compared to what?</strong></p><p>There is no answer key for a plan. And that isn't because AI made the problem hard. It's because we never wrote one.</p><p>This matters more than it sounds. Every evaluation conversation in this space eventually arrives at accuracy, and accuracy stalls immediately, because emergency management outputs describe futures that haven't happened. When a practitioner hears "accurate" with no number and no context attached, they assume total. Nearly every accuracy claim in this market is currently unfalsifiable — which, incidentally, punishes the vendors doing the harder work, because rigor is invisible when nothing is measured.</p><p>Which means that right now, the real measure is persuasion.</p><p>Any vendor can claim accuracy. Any vendor can claim safety. There is no common definition of either in this domain and no standard to be held against — so what actually decides procurements is relationships, sales skill, and marketing. Who you know. Who presented well. Who the emergency manager in the next county already bought from.</p><p>I want to be clear that I'm describing a market I'm in, not one I'm standing above. I cannot prove my own accuracy claims to the standard I'm asking for. Nobody can, because the standard doesn't exist. That is the entire problem.</p><p>So here's what I think happens next, and I'd rather say it before than after: parts of this field are going to get burned by what they are adopting right now. Not because the technology is bad, and not because anyone acted in bad faith. Because we are making high-consequence purchases on charisma, and some meaningful fraction of those decisions will turn out badly, in public, at the worst possible moment.</p><p>When that happens, the damage won't stay with the jurisdiction that bought it. It will land on the whole idea of AI in emergency management, and it will set back the people doing careful work by years. That is the real cost of an unfalsifiable market. Standards aren't a bureaucratic nicety here — they're how a field protects itself from its own enthusiasm, and how the vendors doing rigorous work get to prove it.</p><p>But the accuracy problem is still downstream. The root is that the field never built the habit of evaluating its own work against standards, long before any of this was automated.</p><p>Worth noticing: in the report's own tally of promising AI use cases, developing after-action reviews sits near the bottom. We are asking AI to help with nearly everything except the mechanism that would tell us whether any of it worked.</p><h2 id="what-that-means-for-adding-ai">What that means for adding AI</h2><p>Layering AI onto a field that runs on good vibes and requirements language does not make the field clearer. It makes it faster and more voluminous.</p><p>We will produce more documents, more quickly, still marked complete, with no better idea whether any of them work. The compliance artifact was never the point — a plan is a lossy compression of operational reality, a receipt for thinking that happened somewhere else. Automating the production of receipts makes them cheaper. It does not make a jurisdiction more prepared, and it can make things worse by making the receipt look more complete than the work behind it.</p><p>This is my honest worry about the next three years, and it isn't a worry about the technology. It's that we'll industrialize an unmeasured process and call the throughput progress.</p><h2 id="the-picture-the-report-doesn-t-try-to-paint">The picture the report doesn't try to paint</h2><p>One more thing. The report is an accurate picture of where we are, and it doesn't try to be a picture of where we're going. We need both, and the second one is harder.</p><p>Here's my version.</p><p>Software in this field will change more in the next few years than it has in the last twenty. Practitioners will stop buying an application and adapting themselves to it. They'll have agents build what they need, when they need it, shaped to their jurisdiction and their hazards. The interface stops being the product.</p><p>Which means the durable value moves down the stack, to services and infrastructure. The hard parts. The structured data, the operational context, the domain logic and the guardrails that an agent cannot conjure for a user out of nothing. Log in or don't — the point is that the hard bits should be available to you either way, wherever you and your agents decide to do the work.</p><p>I think that's good for the field, and I think vendors who are honest with themselves know it's coming.</p><h2 id="why-the-knowledge-layer-is-the-durable-one">Why the knowledge layer is the durable one</h2><p>And that is exactly why the missing foundation matters more, not less.</p><p>If agents are going to build against something, this field had better have written down what it knows. What the terms mean. What good looks like. What counts as evidence. What separates a capability that exists from a capability that was discussed.</p><p>Otherwise we've automated the production of confident noise at a scale we've never seen — and we still won't have an answer key.</p><p>Right now we have 1,179 products and no dictionary. No agreed definition of accuracy for this domain. No competency framework an accredited program could teach from. No standard of evidence a clearinghouse could measure against. A capacity-building agenda is the right response to the adoption gap, and I'll support every part of it. But we should be honest that we're building the field's capacity to adopt AI into a discipline that hasn't finished defining itself.</p><p>Every profession that matured did it the same way. It wrote down what it knows — definitions, competencies, standards of evidence — somewhere durable enough to cite and argue with. Emergency management has that. AI in emergency management does not.</p><p>What this field needs is an independent body of work, and I mean each word.</p><p><strong>Independent of any single vendor, including mine.</strong> The moment a company's proprietary framework becomes the field's vocabulary, the vocabulary stops being neutral and starts being a moat.</p><p><strong>Independent of any one initiative or funding cycle.</strong> Programs end. Foundations change priorities. A profession's knowledge has to outlive whoever paid for the first version of it.</p><p><strong>And genuinely a body of work</strong> — peer-reviewed, citable, and open to being argued with. That last part matters most. A body of knowledge earns authority by being contestable. Anything that can't be challenged in public isn't a standard, it's a marketing claim with better formatting.</p><p>Built by practitioners, researchers, and educators — not by the people selling into the field. I say that as one of the people selling into the field.</p><p>That is what I want to see exist, and what I'd want to help build. Some of that work is underway. I'll have more to say before long.</p><p>---</p><p>If you're at the AIDE Summit in Austin next week, find me — I'd rather have this conversation in person.</p><p>And if you're teaching this, researching it, or deploying it in a real jurisdiction: tell me what belongs in it. I'm collecting.</p><p><em>Justin Snair is the founder and CEO of Preppr.</em></p>]]></content:encoded>
</item>
<item>
<title>The Rise of PrepOps</title>
<link>https://blog.preppr.ai/a/the-rise-of-prepops/</link>
<guid isPermaLink="true">https://blog.preppr.ai/a/the-rise-of-prepops/</guid>
<pubDate>Tue, 07 Jul 2026 12:00:00 GMT</pubDate>
<dc:creator>Justin Snair</dc:creator>
<category>Industry Thought</category>
<description>For twenty years we've run preparedness as a project — a plan you finish, an exercise you survive, a report you file. PrepOps runs it as an operation: continuous, grounded, and alive. Here's what that means, and why it's a genuinely new idea.</description>
<content:encoded><![CDATA[<p>For twenty years we've run preparedness as a project — a plan you finish, an exercise you survive, a report you file. PrepOps runs it as an operation: continuous, grounded, and alive. Here's what that means, and why it's a genuinely new idea.</p><p>Every mature discipline eventually discovers the same thing: the work is never "done." Software learned it. Security learned it. Revenue learned it. Preparedness is a serious function still pretending otherwise.</p><p>We still talk about preparedness the way we talked about shipping software in 2005 — as a release. You spend months building the plan. You freeze it. You test it once a year in a room full of people who'd rather be somewhere else. You write the after-action report. You file it. And then, quietly, the world keeps moving while the plan sits still.</p><p>That gap — between a static plan and a living operation — is where preparedness goes to die. PrepOps is the name for closing it.</p><h2 id="what-ops-actually-means">What "Ops" Actually Means</h2><p>Before PrepOps sounds like another buzzword, it's worth being precise about what the "Ops" suffix has meant everywhere it's shown up — because it's meant the same thing every time.</p><p>When software teams adopted DevOps, they weren't buying a tool. They were rejecting a premise: that "build" and "run" are separate phases handled by separate people at separate times. They collapsed the two into one continuous loop. Ship small, observe, correct, ship again. The release stopped being an event and became a heartbeat.</p><p>Security did the same thing with SecOps. Revenue teams did it with RevOps. In each case the move was identical: take something that used to be episodic, periodic, and owned by a silo — and make it continuous, instrumented, measured, and shared. The artifact stopped being the point. The work flow became the point.</p><p>PrepOps is that same move, applied to preparedness. It takes the episodic rituals of emergency management — the plan, the exercise, the assessment, the improvement cycle — and turns them into one continuous operating loop that runs at the speed the threat environment actually changes.</p><h2 id="why-preparedness-got-stuck">Why Preparedness Got Stuck</h2><p>Preparedness didn't stay episodic because emergency managers are behind the times. It stayed episodic because the tools forced it to.</p><p>A plan is a document, so it behaves like a document: authored slowly, approved formally, and immediately out of date. An exercise is a production, so it behaves like a production: scheduled quarters in advance, staffed by a committee, performed once, and remembered mainly by the people who wrote it. An after-action report is a deliverable, so it behaves like a deliverable: completed, submitted, and shelved.</p><p>None of these things are wrong. They're just slow — and worse, they're disconnected. The plan doesn't know what the exercise revealed. The exercise doesn't know what changed in the threat picture last week. The after-action report doesn't feed back into the plan except through the heroic memory of one coordinator. Each artifact is a snapshot, and snapshots start decaying the moment they're taken.</p><p>So preparedness lives in a permanent lag. New staff arrive, new hazards emerge, authorities shift, a vendor changes, a road closes — and the "current" plan quietly describes a world that no longer exists. We've all seen the tabletop where the first thirty minutes are spent arguing over which version of the contact list is real. That's not a preparedness problem. That's a tempo problem.</p><h2 id="prepops-defined">PrepOps, Defined</h2><p>PrepOps is preparedness run as a continuous operation rather than a series of projects. Concretely, it means four things working together.</p><ul><li><strong>The plan is a live system, not a frozen file.</strong> It updates as your reality updates — new documents, new data, new decisions — instead of waiting for the annual revision cycle.</li><li><strong>Exercises become a cadence, not an event.</strong> You test against the world as it is now, whenever the world changes, rather than testing last year's assumptions on a date that suited everyone's calendar.</li><li><strong>Findings loop back automatically.</strong> A gap surfaced in an exercise, an assessment, or a real incident becomes an input to the plan — not a line item in a PDF nobody reopens.</li><li><strong>Everything is grounded and cited.</strong> Because the loop moves fast, every claim has to be traceable — back to doctrine, back to your own commitments, back to the data. Speed without grounding is just confident guessing.</li></ul><p>Put simply: PrepOps is preparedness work that moves as fast — and stays as alive — as operations.</p><h2 id="why-this-is-genuinely-new">Why This Is Genuinely New</h2><p>It's fair to be skeptical. Emergency management has heard "continuous improvement" for decades. The Preparedness Cycle already draws a nice loop on the slide: plan, organize, train, exercise, evaluate, improve. So what's actually different?</p><p>The difference is that the loop was always drawn but never ran. It ran at the speed of meetings — because that was the only speed available. Nobody had a way to keep a plan, a threat feed, an exercise, and a set of findings synchronized in anything close to real time. The connective tissue didn't exist, so the cycle degraded into a sequence of disconnected projects wearing the costume of a loop.</p><p>Two things changed. First, the doctrine, hazard science, and best practice that preparedness rests on became something a machine can hold, reason over, and enforce — deterministically, not by vibe. Second, AI made it possible to read your plans, your uploads, and your day-to-day work and continuously check them against that foundation. For the first time, the loop can run on its own between meetings instead of waiting for the next one.</p><p>That's the new part. Not a new framework — a new tempo. PrepOps is what the Preparedness Cycle becomes when it finally has an engine underneath it.</p><h2 id="what-it-feels-like-in-practice">What It Feels Like in Practice</h2><p>The test of any "Ops" shift isn't the diagram — it's the daily experience. Here's the difference in felt terms.</p><p>In the old model, you open a plan when something forces you to: an audit, a grant deadline, an incident. The rest of the year it's inert. Work happens in bursts of panic separated by long stretches of drift.</p><p>In the PrepOps model, the platform is a workspace you actually work in — drafting, asking, exercising, coordinating. And because you're working in it, it's learning. Every 24 hours it tells you what it noticed: a plan that now contradicts a capability you documented, a hazard assumption that no longer matches the live feed, a pattern in how your team actually makes decisions. You accept what's right, dismiss what isn't, and the system sharpens. The burden isn't a quarterly scramble. It's a few minutes of review against a system that's been paying attention on your behalf.</p><p>Preparedness stops being evidence you reconstruct before a deadline and becomes something you maintain continuously — the way an operations team maintains a service.</p><h2 id="the-objection-worth-taking-seriously">The Objection Worth Taking Seriously</h2><p>Speed is dangerous in this domain. A software deploy that goes sideways triggers a rollback. A preparedness decision that goes sideways can cost lives downstream. So the obvious worry about making preparedness "fast" is that fast means sloppy.</p><p>That worry is exactly right, and it's why PrepOps only works on top of a deterministic, grounded foundation. The reason a modern operations team can move fast without breaking things is that every change runs against tests, guardrails, and observability.</p><p>PrepOps needs the same discipline: a base of rules derived from real doctrine and regulation, hazard models tied to real data, and citations on every work product so a human can check the machine's homework.</p><p>The point of PrepOps isn't to replace judgment with speed. It's to remove the drudgery that was stealing time from judgment — the version-hunting, the reformatting, the reconstruction — so that practitioners spend their attention on the decisions only they can make.</p><h2 id="where-preppr-fits">Where Preppr Fits</h2><p>This is the category we're building Preppr for. Underneath the products sits a deterministic foundation — a large body of doctrine, best practice, and regulation compiled into rules, plus hazard models and infrastructure data. On top of that sits your account norms: your plans, your uploads, your connected data, analyzed for the commitments they make. And running across both is a learning layer that watches how you actually work and surfaces what's drifting, what's inconsistent, and what you've implicitly decided.</p><p>The AI products — content and gap analysis, threat intelligence, stakeholder engagement, exercise and simulation design and delivery, improvement — are the tools you reach for. But the reason they add up to PrepOps rather than just "more software" is that they all feed one system that compounds. Use the workspace, and preparedness stops resetting to zero after every exercise. It accrues.</p><h2 id="the-shift-is-already-underway">The Shift Is Already Underway</h2><p>Disciplines don't adopt an "Ops" mindset because a vendor names it. They adopt it because the old tempo stops being survivable. Software couldn't ship yearly once the internet made change constant. Security couldn't audit annually once attackers moved daily. Preparedness is hitting the same wall: threats compound faster than the planning cycle can absorb, and the gap is widening.</p><p>PrepOps is simply the honest response to that reality — preparedness that runs continuously because the world it's preparing for does. The organizations that make the shift first won't just be more efficient. They'll be operating in a different tense than everyone else: present, not past.</p><p>Preparedness has always been about the future. It's time the work itself stopped living in the past.</p><p>Want to see what preparedness looks like when it runs like an operation? <a href="https://preppr.ai">Start with Preppr</a>.</p>]]></content:encoded>
</item>
<item>
<title>Time Is the Killer: Why AI Startups and Incumbents in EM Will Partner</title>
<link>https://blog.preppr.ai/a/time-is-the-killer-why-ai-startups-and-incumbents-in-em-will-partner/</link>
<guid isPermaLink="true">https://blog.preppr.ai/a/time-is-the-killer-why-ai-startups-and-incumbents-in-em-will-partner/</guid>
<pubDate>Tue, 07 Jul 2026 12:00:00 GMT</pubDate>
<dc:creator>Justin Snair</dc:creator>
<category>Industry Thought</category>
<description>There is a pattern forming in enterprise and government emergency management and business continuity technology. AI-native startups and established incumbents are pairing up. Not acquiring. Not competing to the death. Partnering.</description>
<content:encoded><![CDATA[<p>There is a pattern forming in enterprise and government emergency management and business continuity technology. AI-native startups and established incumbents are pairing up. Not acquiring. Not competing to the death. Partnering.</p><p>People read this as a story about capability. The startup has the clever product. The incumbent has the money. Each grabs what it lacks.</p><p>That read is too shallow. The real currency is time. And time is the killer for both sides, in opposite ways.</p><h2 id="the-startup-s-problem-is-not-the-product">The startup's problem is not the product</h2><p>Say you have an outstanding product. Genuinely great. Sharp team, real technology, a thing the market has never seen. Made faster than ever with the help of AI.</p><p>That is not the unlock.</p><p>Selling into government and the large private sector is a different puzzle than building. Procurement is its own discipline. It has its own gatekeepers, its own cycles, its own logic that has nothing to do with how good your software is.</p><p>→ Your product only breaks through the bureaucratic slog if it solves a massive, screaming pain. Absent that, quality alone does not move the queue.</p><p>→ Everything else waits. And while you wait, time runs.</p><p>Time is what kills the startup. Not the competition. Not the roadmap. The gap between "we built it" and "someone paid for it."</p><p>You survive that gap one of two ways. Capital in the bank, or another source of runway. If you have neither, the clock wins before the market ever gets to say yes.</p><p>This is where an incumbent partner changes the equation. They do not just bring money. They bring time.</p><p>→ Capital extends your runway, so the clock runs slower.</p><p>→ Their procurement position shrinks the runway you need, because they have already fought the battles you are about to face. The vendor approvals. The security reviews. The contract vehicles. The relationships. Years of slog, already done.</p><p>The incumbent hands the startup time on both ends. More of it, and less of it needed.</p><h2 id="the-incumbent-has-the-same-enemy-wearing-a-different-coat">The incumbent has the same enemy, wearing a different coat</h2><p>Now flip it.</p><p>The incumbent has what the startup dreams about. Resources. An installed customer base. Procurement solved. Distribution that took decades to build.</p><p>And they still lose to time.</p><p>→ Internal bureaucracy is the tax on scale. The bigger the organization, the slower it turns.</p><p>→ Emerging markets do not wait for a large company's planning cycle. By the time the machine decides to move, the window has narrowed.</p><p>The incumbent's problem is not capability or reach. It is velocity. They cannot move fast enough to capture what is emerging, because everything that makes them powerful also makes them slow.</p><p>So it is the same enemy. Time. The startup has speed and no runway. The incumbent has runway and no speed.</p><h2 id="best-of-both-worlds-is-really-a-trade-in-time">Best of both worlds is really a trade in time</h2><p>This is why the partnerships are forming, and why they will keep forming.</p><p>Strip away the language about synergy and platforms. What actually changes hands is time.</p><p>→ The incumbent gives the startup runway and a shortcut through procurement.</p><p>→ The startup gives the incumbent speed and a foothold in a market its own cycle would have missed.</p><p>Each side hands the other the exact thing time was stealing from it. That is the trade. Not product for money. Time for time.</p><h2 id="what-this-means-going-forward">What this means going forward</h2><p>The convergence of AI-native builders and established players is not a fad, and it is not a fire sale. It is a rational response to a shared constraint.</p><p>If you are a startup, the lesson is that a great product is table stakes, not a strategy. The question is whether you can survive the distance between building and selling. Someone who can compress that distance is worth more than someone who can only fund it.</p><p>If you are an incumbent, the lesson is that your scale is not free. It buys you reach and costs you speed. Partnering with those who move fast is not charity to a smaller company. It is how you stay in a market that will not wait for you.</p><p>The winners on both sides will be the ones who understand what they are actually negotiating for.</p><p>It was never the product. It was never the money.</p><p>It was always time.</p>]]></content:encoded>
</item>
<item>
<title>From exercises to an operation that runs</title>
<link>https://blog.preppr.ai/a/operation-that-runs/</link>
<guid isPermaLink="true">https://blog.preppr.ai/a/operation-that-runs/</guid>
<pubDate>Fri, 12 Jun 2026 12:00:00 GMT</pubDate>
<dc:creator>Justin Snair</dc:creator>
<category>Product Update</category>
<description>We started as the complete exercise platform. We've become something larger — a place where your whole preparedness operation lives, learns, and works on its own. Here's what we have now, and what's coming next.</description>
<content:encoded><![CDATA[<p>Preparedness has always been treated as a series of projects. An exercise this quarter. A plan refresh next year. A THIRA when it's due. Each one starts from a blank page, or at best a document to jump off from, ends in a document, and the document goes in a drawer. The next project starts cold.</p><p>We built Preppr because that model wastes the most valuable thing an emergency manager produces: what they learn. Every exercise surfaces gaps. Every after-action names fixes. Every plan encodes hard-won judgment. And almost none of it carries forward. Your readiness, in effect, resets to zero the moment the report is filed.</p><p>So we set out to change the unit of work — from the project to the operation. Not a faster way to write documents, but a system that captures the complexity of what you actually run, and compounds it.</p><blockquote>Preparedness isn't a project with an end date. It's a continuous operation. Software should treat it that way.</blockquote><h2 id="where-we-are-now">Where we are now</h2><h3>A preparedness operations platform</h3><p>What began as exercise design has grown into a connected system. The exercise is still there — and it's still the best in the category — but it's now one motion inside a platform that holds the whole lifecycle. Three properties define it.</p><p><strong>01 · Connected</strong> — <em>One operation, not ten tools.</em> Assess, gather intel, engage stakeholders, design, exercise, improve — every stage in one workspace, sharing one memory. The work stops living in scattered files and disconnected apps.</p><p><strong>02 · Compounding</strong> — <em>Readiness that builds on itself.</em> Every exercise, finding, and decision flows back into one living knowledge base. The next question starts smarter than the last. Nothing resets to zero.</p><p><strong>03 · Grounded</strong> — <em>Built on what you're held to.</em> Hundreds of rules across every sector you operate in, plus your own plans and patterns. Every output cites the requirement it meets — defensible, not generic.</p><table><thead><tr><th>~1,500</th><th>Rules from federal doctrine &amp; regulation</th></tr></thead><tbody><tr><td>14+</td><td>Sectors covered out of the box</td></tr><tr><td>5</td><td>Lifecycle stages, one connected system</td></tr><tr><td>∞</td><td>Findings that compound, never reset</td></tr></tbody></table><p>On top of that foundation sit the capabilities people use every day. <strong>Ask Preppr</strong> reasons across your knowledge base to develop plans and answer hard questions. <strong>Collaborate</strong> brings external stakeholders into the work with real roles. The <strong>Exercise Designer</strong> turns design from a content problem into an agentic build — it plans the work, reasons across your doctrine and history, and hands back a defensible exercise. <strong>Preppr Exercise</strong> then takes it live — delivering and facilitating the exercise itself, and feeding what surfaces straight back into the record. And the <strong>Knowledge Base</strong> ties it all together, linking plans, exercises, findings, and rules into a single graph.</p><h2 id="the-shift">The shift</h2><h3>From a tool you use to an operation that runs</h3><p>The biggest change isn't a feature. It's a change in posture. Most preparedness software is a place you go — you open it, you do work, you leave. Preppr is increasingly an operation that runs.</p><p>This is worth being precise about, because the words for it have been worn smooth. "Connected" has come to mean a vendor's software can reach into your tools — an integration, a plumbing claim. We mean the opposite. You never come to Preppr. You stay in your inbox, your Slack, your meeting, and the same operation runs whether you typed it, said it, or forwarded it from your phone. Not <em>we can call your tools</em> — <em>you never have to leave where you already work.</em> That's a different axis entirely, and it's the one that matters when the work competes with a full calendar.</p><p>"Agent" has been worn smoother still. For most software it means a chatbot reading over your documents — it answers questions about your plan. Preppr does something a chatbot structurally cannot: it pressure-tests the plan. Hand it a tabletop and it doesn't summarize your doctrine back to you — it builds the exercise, runs it dozens of ways, finds where it breaks, and hardens it before you ever see a draft. Answering questions is retrieval. Doing the analysis you couldn't run by hand is labor. Only one of them changes how ready you are.</p><p>You direct it however you already work: by typing or clicks, by voice, by handing it a goal over email or Slack, or by driving it from your own tools. It does the multi-step work in the background and reports back where you asked. The most reliable member of your team — one you delegate to, not one you operate.</p><p>And there's a hard line we hold: <strong>autonomy on the labor, never on the decisions.</strong> Preppr does the building and the analysis a person could never do by hand — but it pauses before anything goes outward, and nothing reaches your leadership until you say yes. You keep the judgment. It takes the work. And because the analysis runs out of sight, you can always look in: the full fleet of work behind a recommendation is open to you, including what it tried and set aside, not just the answer it landed on.</p><blockquote>You supply the intent and the judgment. Preppr supplies the plan and the labor. The line between them never blurs without your say-so.</blockquote><h2 id="what-s-soon-possible">What's soon possible</h2><h3>Where this goes next</h3><p>Everything above is live or arriving. What makes us most optimistic is what the foundation now makes possible — because once the operation is connected, compounding, and grounded, entirely new capabilities become straightforward.</p><p><strong>→ Autopilot, end to end.</strong> Hand Preppr an outcome — "draft a hurricane tabletop for the EOC and schedule it next month" — and it designs, validates, asks only what it can't infer, pauses for your approval, then finishes and reports back. It confirms the request really came from you, works in the background, and you never open the app unless you want to look in.</p><p><strong>→ Simulate at scale.</strong> Before a design ever reaches you, Preppr fans out a cohort of concurrent agentic simulations — each a full multi-cell tabletop in miniature, grounded by a deterministic physics engine and a cognitive model so every agent reasons like a different person under pressure. A Red cell drives the threat, a White cell adjudicates, a Blue cell of section chiefs deliberates — and they disagree, because each seat carries its own expertise, risk appetite, and priorities. Preppr reads across the whole fleet and hardens the exercise before a human sees it. It's a depth of analysis a chatbot over your documents can't reach, and a person could never run by hand.</p><p><strong>→ Intelligence that informs every decision.</strong> The point was never just to produce preparedness intelligence — it's to put it to work. As the knowledge base deepens, Preppr surfaces what matters across the whole cycle, so every decision is better-informed than the last, not just the documents faster.</p><p><strong>→ Work anywhere.</strong> Talk to it. Email it. Message it. Or let it run. Preppr meets you in the channels and tools you already use, so the operation keeps moving whether or not you're sitting in front of it. Not connected to your tools — running from inside the way you already work.</p><p><strong>★ Anticipatory readiness — Preppr watches, then acts.</strong> This is where it all converges. Preppr monitors the threats and trends you named in your THIRA, HVA, and risk assessment. When one moves, it doesn't wait to be asked — it reads your knowledge base, sees what you've worked on and where the gaps still are, and acts: a heads-up by email or Teams, a plan offered before you'd have thought to build it, and recommendations grounded in what Simulate already learned about how that scenario plays out against your operation. The work shifts from reacting to a calendar to staying ahead of the risk itself.</p><p>The throughline is simple. You describe the outcome. Preppr does the work — grounded in your reality, compounding with every cycle, and always stopping at the decisions that are yours. The result is the thing every emergency manager is actually trying to be:</p><p><strong>The most prepared person in the room — and getting more so every cycle, without the busywork that used to be the job.</strong></p>]]></content:encoded>
</item>
<item>
<title>Why Your Preparedness Resets to Zero After Every Exercise</title>
<link>https://blog.preppr.ai/a/why-your-preparedness-resets-to-zero-after-every-exercise/</link>
<guid isPermaLink="true">https://blog.preppr.ai/a/why-your-preparedness-resets-to-zero-after-every-exercise/</guid>
<pubDate>Wed, 10 Jun 2026 12:00:00 GMT</pubDate>
<dc:creator>Justin Snair</dc:creator>
<category>General</category>
<description>You ran a good tabletop. The right people were in the room. The scenario was sharp, the conversation was honest, and the after-action report named real gaps. People left…</description>
<content:encoded><![CDATA[<p>You ran a good tabletop. The right people were in the room. The scenario was sharp, the conversation was honest, and the after-action report named real gaps. People left saying it was the most useful day they'd spent on preparedness all year.</p><p>Six months later, almost none of it exists.</p><p>The AAR is a PDF in a shared drive. The improvement items got half-tracked in a spreadsheet, then overtaken by the next emergency. The person who understood the hospital surge dependency took a new job. The next exercise will start, more or less, from a blank page — re-litigating questions you already answered, re-discovering gaps you already found.</p><p>This isn't a failure of effort. It's the most common pattern in emergency management, business continuity, and crisis work, and the people running these programs are not lazy or unskilled. They're some of the most capable operators in any field. The problem isn't the practitioners. It's the process they're forced to work inside — a process that is <em>designed</em> to throw away most of what they know.</p><h2 id="preparedness-is-a-matching-problem">Preparedness is a matching problem</h2><p>There's a useful idea from cybernetics that explains why this happens, and it's worth the two minutes it takes to understand.</p><p>In 1956, the British cyberneticist W. Ross Ashby formulated what became known as the <strong>Law of Requisite Variety</strong>: a system can stay in control of its environment only if its internal variety is at least as large as the variety of the disturbances coming at it. In his phrasing, <em>only variety can absorb variety.</em></p><p>"Variety" here doesn't mean assortment. It means the number of meaningfully different situations a system can actually tell apart and respond to — its real repertoire of distinct, usable responses.</p><p>Apply that to preparedness and the whole field comes into focus. A threat environment has enormous variety: a near-infinite number of ways things can go wrong, compounding and recombining, throwing up novel situations no one scripted. Your response system's variety is how many of those you can actually recognize and act on. <strong>Preparedness is the match between the two.</strong></p><p>Ashby's Law says something uncomfortable about that match. When the threat environment generates vast, compounding, novel variety, and the response apparatus is thin, static, and built around a handful of canned plans, the gap isn't bad luck. It's failure by definition. The controller is simply outmatched by what it's trying to control.</p><p>So the real question for any preparedness program isn't "do we have a plan?" It's "does our repertoire of responses come anywhere close to the variety of what could actually happen to us?" For most organizations, honestly answered, the answer is no — and not because they haven't worked hard at it.</p><h2 id="the-process-is-lossy-by-design">The process is lossy by design</h2><p>Here's the part that reframes everything.</p><p>The way preparedness usually gets done isn't just under-resourced. It actively destroys variety. It takes a high-resolution operating reality — varied, local, full of edge cases and weak signals — and runs it through a process whose real output is a compliance artifact: the plan, the AAR, the signed exercise documentation.</p><p>To produce that artifact on time, on template, and on budget, the process has to compress. Hard. Distinct hazards collapse into generic categories. Local nuance becomes boilerplate. Edge cases and weak signals — exactly the things that matter most when something novel hits — get deleted, because they don't fit the form. What comes out the other end is thin, generic, and frozen the moment it's filed.</p><p>The receipt becomes the product. And the variety that <em>is</em> preparedness gets destroyed on the way to the filing cabinet.</p><p>It's worth sitting with how backwards this is. Practitioners don't lack variety — they carry an enormous amount of it in their heads, their relationships, and their judgment. The process strips it out. We've built an entire discipline around manufacturing certificates that preparedness happened, and in doing so we routinely throw away the thing that would have made us prepared.</p><h2 id="faster-documents-makes-it-worse-not-better">"Faster documents" makes it worse, not better</h2><p>This is also why a lot of the new AI tooling in our space is solving the wrong problem.</p><p>Most of it is, at bottom, a faster way to generate documents. Point it at a prompt and it produces a plausible plan, a plausible exercise, a plausible AAR. That feels like progress because the painful part of the old process — the blank page, the hours of drafting — gets faster.</p><p>But look at what it's actually accelerating. It's the <em>compression</em> step. It's the receipt factory, running quicker. More documents, produced faster, frozen just as solid the moment they're done. A confident, plausible-sounding document with no model of your actual reality underneath it isn't requisite variety. It's the same lossy output with a shorter turnaround.</p><p>Speeding up the thing that was destroying variety doesn't close the gap. It widens it.</p><h2 id="what-actually-closes-the-gap">What actually closes the gap</h2><p>The fix isn't more content. It's raising your response variety so it comes closer to matching your real threat environment — and, critically, keeping that variety alive instead of compressing it away after every cycle.</p><p>Concretely, that means a few things working together:</p><ul><li><strong>Preparedness that compounds rather than resets.</strong> Every exercise, plan, decision, and finding should make the next one sharper — not vanish into a drive. A tabletop you run shouldn't be an event you certify and forget; it should be a deposit into a system that remembers. The variety you generated this quarter should still be there, and usable, next quarter.</li><li><strong>Grounding in real doctrine and live conditions.</strong> Response variety is only useful if it's true. That means answers anchored in actual federal doctrine and live hazard data — with citations and an audit trail — not a model guessing confidently in a setting where being wrong has consequences.</li><li><strong>An information architecture that matches the complexity it's facing.</strong> One connected system that can see across your exercises, plans, intelligence, and stakeholders — and model how a failure in one place cascades into another, the way an energy outage cascades into hospital staffing, or a water failure into a public-health surge. Single-vertical, single-document tools structurally can't see those cascades. The real world doesn't respect those boundaries, so a system meant to match it can't either.</li></ul><p>This is the conviction Preppr is built on. Not "more content, faster" — that's just faster receipts. The goal is to give practitioners enough usable, grounded, connected response variety to actually meet the environment they operate in, and to make sure that variety <em>accumulates</em> instead of evaporating after every cycle.</p><p>Only variety absorbs variety. The job is to raise the line.</p><h2 id="the-test-worth-applying">The test worth applying</h2><p>You don't need a product to ask the question that matters. Ask it of whatever you're using now:</p><p><em><strong>When this exercise/analysis/threat assessment is over, will we be measurably better when we do the next one — or will we be starting over?</strong></em></p><p>If the honest answer is "starting over," it doesn't mean your team failed. It means the process did exactly what it was designed to do: produce a receipt, and let the variety walk out the door.</p><p>The opportunity in front of our entire field is to stop confusing the certificate with the capability. Preparedness shouldn't reset to zero. It should compound.</p>]]></content:encoded>
</item>
<item>
<title>Truth, Story, and the Loop Between Them: How Preppr's New Engine Builds Preparedness Work You Can Actually Trust</title>
<link>https://blog.preppr.ai/a/truth-story-and-the-loop-between-them/</link>
<guid isPermaLink="true">https://blog.preppr.ai/a/truth-story-and-the-loop-between-them/</guid>
<pubDate>Sat, 23 May 2026 12:00:00 GMT</pubDate>
<dc:creator>Justin Snair</dc:creator>
<category>General</category>
<description>There's a problem with using AI to do serious preparedness work, and anyone who's used AI long enough has run into it. The language is good. The facts aren't — or at least…</description>
<content:encoded><![CDATA[<p>There's a problem with using AI to do serious preparedness work, and anyone who's used AI long enough has run into it. The language is good. The facts aren't — or at least, not always, and you can't tell when.</p><p>An AI can draft a hurricane scenario for a hospital that sounds exactly like the real thing — the cadence of incident command, the texture of public messaging, the right acronyms in the right places. But ask it whether the configured population in the scenario matches the actual county. Ask it which CMS regulations are binding for that facility. Ask it what realistically happens to fuel reserves over 72 hours of generator load. The AI will answer all three confidently. Two of those answers will be wrong, and you won't be able to tell which two.</p><p>For a casual brainstorm, that's a minor annoyance. For a compliance exercise, an after-action review, or training that informs real plans, it's disqualifying.</p><p>Preppr is built around a different architecture, and it's worth describing plainly because it changes what AI in this domain can actually be trusted to do.</p><h2 id="the-engine-is-the-source-of-truth">The Engine Is the Source of Truth</h2><p>At the center of Preppr is a new piece of software we call the Compliance &amp; Operational Physics Engine. It's not a chatbot. It's a deterministic simulator — software that, given a description of an entity under stress (a single facility, an organization, a jurisdiction, or a whole community) and a hazard, runs it forward through time according to rules and equations.</p><p>The engine carries two kinds of authoritative ground truth — rules and data — and runs two kinds of "physics" against them, side by side, over a shared simulated world.</p><p><strong>Compliance physics.</strong> The full force of federal, state, accreditation, and sector-specific rules — currently 489 and growing, across 18 sectors — evaluate continuously against that world. When fuel drops below the critical threshold, the CISA notification deadline starts. When water pressure falls, EPA's Tier 1 public-notice clock activates. When a federal declaration arrives, a different set of grant-eligible obligations turns on. Every rule is tied to a real citation — CMS 482.41, AWIA 2013, 44 CFR 206, NFPA 1600 — not a paraphrase, not a guess.</p><p><strong>Operational physics.</strong> What actually happens during the incident. Hazards onset and worsen. Fuel and water and staffing burn down at sector-appropriate rates. People get fatigued. The public surges into the ER. Power loss cascades into communications degradation cascades into command disruption. Impact gets weighted by the real social-vulnerability profile of the affected population, drawn from authoritative sources the engine grounds against — FEMA's National Risk Index across all 3,232 U.S. counties and 56 states for vulnerability and exposure, FEMA Hazus for earthquake loss baselines, curated outbreak signal for emerging infectious disease where the engine has it. The list of registered sources grows as the engine takes on new hazards and regions.</p><p>Re-run the same scenario with the same inputs and you get a bit-for-bit identical audit log. Every event in the log carries its rule id, citation, and provenance hash. That's the difference between a defensible answer and a confident-sounding one.</p><h2 id="the-language-models-write-the-story">The Language Models Write the Story</h2><p>The engine, on its own, is austere. It tells you what's true, not what's interesting. That's where language models come in — and where most of the user experience lives.</p><p>When you build an exercise in Preppr's Exercise Designer, language models draft scenarios and write injects or pull on hazard mitigation plans, and frame the questions players will face. When you talk to Ask Preppr, language models compose answers, reason across your knowledge, and synthesize analysis. When you generate an alert template, an after-action report, or an EEG, language models do the writing.</p><p>Language models are good at this. They're good at language, at opinionated judgment, at framing, at synthesis. They can take a dry sequence of regulatory triggers and turn it into a scenario your partners will actually engage with. They can take the engine's terse "fuel reserves crossed critical threshold at T+47:00" and turn it into a narrative your leadership will respond to.</p><p>What they're not good at is keeping the truth straight, the rules current, or the cascades physically consistent. So Preppr doesn't ask them to.</p><h2 id="then-the-engine-checks-the-work">Then the Engine Checks the Work</h2><p>This is the part that matters. When a language-model output touches the live world, the engine validates it against the truth.</p><p>When Preppr's Exercise Designer drafts a scenario, the configuration runs through the engine's plausibility check against the registered authoritative sources for that hazard and jurisdiction. If the configured magnitude is two full steps off the nearest modeled scenario for the county, or if the affected-population number is ten times the displaced-household estimates the data supports, the engine flags it before the scenario is frozen. The sources doing the checking are the engine's authoritative ground truth — NRI and Hazus today, and a growing set of registered feeds, like flood forecasting, UNDRR hazard profiles, and WHO epi data, as new hazards and regions come online.</p><p>When an action proposes citing 44 CFR 206.2, the engine verifies that the citation matches the loaded rule corpus and that the rule actually applies to the facility, sector, and hazard in question. A drafted citation that doesn't bind doesn't get stamped into the artifact.</p><p>When a scenario goes through pre-simulation validation, the engine surfaces which rules are actually on the hook — by jurisdiction scope, by sector, by capability — and reports them with their full citation chain. The language model can describe; the engine confirms.</p><p>When the language model and the engine disagree, the engine is authoritative. That's the contract. It's what keeps the AI-assisted parts of Preppr from drifting on vibes into something that feels right but isn't.</p><h2 id="extending-what-s-true-and-what-s-knowable">Extending What's True — and What's Knowable</h2><p>The engine knows a lot. It doesn't know everything. And the language models, when they write the story, draw on more than just the engine — they also draw on the body of knowledge your organization brings to the table. Preppr is designed so both layers can grow, along two symmetric axes.</p><p>The engine extends through new authoritative ground truth — rules and data. Grounding extends through new bodies of knowledge — personal and shared.</p><p><strong>Workspace data sources extend the engine (operational side).</strong> Your admin can register an external authoritative API — WHO's GREPI for global infectious-disease events, USGS earthquake feeds, your state's emergency-management open data, your sector's threat-intelligence service — as a new authoritative data source for your organization's scenarios. The engine fetches a sample, runs five validation gates (the data has to be present, the hazards have to be engine-known, the field mappings have to map to engine-consumable paths, the publisher has to be trustworthy or backed by an admin attestation), and then refreshes every 24 hours. Approved sources feed scenarios automatically, supplying defaults for jurisdictions FEMA doesn't cover and contributing cascading-hazard probabilities where the federal data has gaps. Federal data still wins on overlap. Every parameter the engine pulls from a registered source carries its origin with it; downstream artifacts trace back, source by source, to the authoritative material they're built on. Three authoritative sources, one coherent scenario, every parameter traceable.</p><p><strong>Personal Knowledge Bases extend what the language models can ground against (private).</strong> Documents you upload, conversations you've had with Ask Preppr, intel briefs you've captured, exercise designs you've created, after-action findings from past work — they all flow into your personal Knowledge Base and become retrievable when you ask questions or design new artifacts. Ask Preppr cites them. Designer references them. They're how the AI knows what your organization has been working on, not just what's in the federal corpus.</p><p><strong>Preppr Knowledge Collections (Packs, in the codebase) extend grounding across people (shared).</strong> A Collection is a curated body of content — doctrine, playbooks, lessons learned, sector best practices — that a team can subscribe to and share. When you turn a Collection on, its content becomes part of what Ask Preppr will cite and what Designer can draw on. Collections are how an organization's institutional knowledge becomes part of every member's AI-assisted work.</p><p>The first pair extends the engine — where truth is enforced, with citations and provenance. The second pair extends grounding — where story is generated, with retrievals that can be traced back to the document they came from. All four are admin-gated where they need to be, validated where validation matters, and structured so end users can't slip unverified content into the parts of the system that have to remain authoritative.</p><h2 id="the-whole-thing-is-addressable">The Whole Thing Is Addressable</h2><p>The engine, the simulator, your personal Knowledge Base, the Preppr Knowledge Collections you toggle on, the exercise designs you've built, the scenarios in your library — all of it is accessible via MCP, the Model Context Protocol. Read and write, both directions.</p><p>That means the AI tools your team already uses — Claude, ChatGPT, Copilot, whatever sits on your analysts' desktops — can query Preppr directly. An emergency manager working in Claude can ask "what would the engine say about a Category 3 hitting our coastal facilities," get a real engine run back with citations, and pull the result into a brief without leaving the chat. A planner in their own tooling can pull a scenario out of Designer, edit it, and write it back. A researcher can query a Collection, retrieve grounded passages, and cite them in an analysis their organization will actually trust.</p><p>This matters for three reasons.</p><p>First, the truth-and-story architecture stops being a Preppr-only architecture. The engine's authoritative outputs — rule firings with citations, plausibility checks against registered sources, resource-burn calculations grounded in equations — can underwrite work happening in any AI surface your organization uses. The deterministic layer becomes a service the rest of your stack can lean on, not a walled garden you have to log into to benefit from.</p><p>Second, your institutional knowledge stops being trapped in the tool that captured it. Personal KBs, Collections, past exercise designs, after-action findings — they remain retrievable, citable, and groundable from wherever your team is actually working. The investment your organization makes in building up its substrate compounds across every AI-assisted workflow it touches, not just the ones that happen inside Preppr's UI.</p><p>Third, and most consequentially: the validation loop extends outward. When an external AI calls the engine through MCP, it gets the same audit log, the same citations, the same deterministic guarantees, and the same provenance trail that an internal call would get. The contract doesn't weaken at the boundary. The story can be written anywhere; the truth still has one source.</p><p>This is what we mean when we say Preppr is a preparedness intelligence layer, not an app. The app is one way in. MCP is how the layer becomes part of everything else.</p><h2 id="why-this-matters">Why This Matters</h2><p>The architecture is, in the end, a stance: about what AI in critical-domain work should be doing and what it shouldn't.</p><p>Language models should do what they're good at — turning truth into story, drafting artifacts people will engage with, synthesizing across a body of knowledge, generating language that lands. They should not be asked to keep the regulatory deadlines current, the resource burn rates physically consistent, or the cascades accurate. Those are tasks for a deterministic engine grounded in real rules and real data, with citations and provenance attached.</p><p>The reason Preppr can produce defensible preparedness artifacts is that the engine sits underneath the whole thing, checking the language models' work at every step. Every rule that fires has a citation. Every resource level is traceable to an equation. Every plausibility warning is grounded in a registered authoritative source. Every retrieval is auditable back to a document in a KB or a Collection.</p><p>For an emergency manager, this means the AI-assisted work you do in Preppr is AI-assisted in the parts where AI is helpful — drafting, framing, narrative, synthesis — and not AI-assisted in the parts where it isn't: rules, numbers, physics, citations. The artifacts you put in front of leadership are grounded in real data and real regulation. The plans you build reflect what would actually apply to your facility, your organization, your jurisdiction, your community.</p><p>Truth and story, in the right roles, with a loop between them. That's the architecture.</p><p>Ready for emergency management AI you can actually trust? <a href="https://preppr.ai/demo">Book a demo.</a></p>]]></content:encoded>
</item>
<item>
<title>Anatomy of a Bad EULA for AI Powered Emergency Management Software</title>
<link>https://blog.preppr.ai/a/anatomy-of-a-bad-eula-for-ai-powered-emergency-management-software/</link>
<guid isPermaLink="true">https://blog.preppr.ai/a/anatomy-of-a-bad-eula-for-ai-powered-emergency-management-software/</guid>
<pubDate>Sun, 17 May 2026 12:00:00 GMT</pubDate>
<dc:creator>Justin Snair</dc:creator>
<category>General</category>
<description>Most emergency management buyers don't read the End User License Agreements (EULA). The procurement officer gives it a look. The IT lead Ctrl-F's for ’SOC 2.’ The program…</description>
<content:encoded><![CDATA[<p>Most emergency management buyers don't read the End User License Agreements (EULA). The procurement officer gives it a look. The IT lead Ctrl-F's for "SOC 2." The program manager wants to know if it talks to WebEOC. Counsel, if counsel is involved at all, looks at the indemnification and the venue clause and calls it a day. Almost no one reads the document end to end.</p><p>I do. I read EULAs the way some people read restaurant menus — looking for what's actually being served, not just what the chalkboard out front says. And in the AI era, the EULA has become the single most diagnostic document a vendor produces. The marketing tells you what the vendor wants you to believe. The demo tells you what works on a good day. The contract tells you what the vendor is actually willing to stand behind when things go wrong.</p><p>In emergency management, "when things go wrong" isn't a hypothetical. It's the entire reason the software exists.</p><p>What follows is an anatomy of a bad EULA in AI/EM software. The clauses I quote are pulled from a real, currently-in-force contract for an AI-powered emergency management platform. I'm not naming the vendor. The clause patterns are what matter, and they're starting to show up across the category like a recognizable cookie-cutter house style — probably coming from prompts like <em>"hey ChatGPT, write me an EULA that keeps our company from being on the hook for everything. K, thanks."</em></p><p>There are seven anatomical features. Each is a red flag on its own. Together they describe a contract that no responsible public agency, healthcare organization, or critical infrastructure operator should sign without significant negotiation — or, in most cases, at all.</p><h2 id="1-the-convenience-only-ai-disclaimer">1. The "Convenience Only" AI Disclaimer</h2><p>The first feature is the clause where the vendor demotes its own AI from a product capability to a perk. From the contract:</p><blockquote><strong>"The Company provides AI Features as a convenience tool only."</strong> <strong>"AI FEATURES ARE EXPERIMENTAL IN NATURE AND ARE OFFERED FOR CONVENIENCE ONLY."</strong></blockquote><p>A <em>convenience</em>. Like a coffee machine in the lobby. Nice to have. Don't sue us if it makes a bad latte. Except the marketing on the way in didn't say "we have a coffee machine." The marketing said "we are THE BEST AI-powered EM platform." Literally the defining feature. This is like a landlord renting out a house on the power grid, but hiding that electricity can be removed whenever the landlord wants. You'd be rightfully livid if your power was turned off.</p><p>So the vendor markets the AI as the product, prices the product as if the AI is the product, and then writes a contract that positions the AI as a non-essential extra. The marketing and the contract are pointed in opposite directions. The buyer is paying for one thing and signing for another.</p><p>There's a useful interpretive trick here: when a contract disclaims something this aggressively, it's telling you what the vendor's own confidence in that thing looks like. A vendor that believed in its AI would warrant it. A vendor that wasn't sure would disclaim it. This vendor isn't sure.</p><p><strong>Question to ask the vendor:</strong> If the AI is offered "for convenience only," what part of the product am I actually paying for? And if the AI is the differentiator in your marketing, why is it disclaimed in your contract?</p><h2 id="2-the-mandatory-human-review-clause">2. The Mandatory Human Review Clause</h2><p>On its face, this is the section that sounds the most defensible. "Human in the loop" is good practice. In high-stakes domains, AI outputs <em>should</em> be reviewed by qualified humans before they're acted on. Any responsible AI EM product — including the one I build — operates under that principle. It's not the principle that's the problem. It's what the clause is doing in this specific contract, in light of what the product actually does.</p><p>From the contract:</p><blockquote><strong>"EVERY OUTPUT GENERATED BY ANY AI FEATURE ... MUST BE INDEPENDENTLY REVIEWED, VERIFIED, AND VALIDATED BY A QUALIFIED HUMAN WITH APPROPRIATE EXPERTISE, AUTHORITY, AND TRAINING BEFORE THE OUTPUT IS APPROVED, ADOPTED, FINALIZED, DISTRIBUTED, ACTED UPON, INCORPORATED INTO ANY PLAN OR PROCEDURE, OR RELIED ON FOR ANY EMERGENCY, CONTINUITY, COMPLIANCE, REGULATORY, LEGAL, OPERATIONAL, OR LIFE-SAFETY DECISION."</strong></blockquote><p>The actual clause runs about 130 words in a single all-caps sentence. The drafting density is its own tell. When a vendor needs that many enumerations and that much capitalization in one sentence, they're not articulating a principle — they're building a wall. The wall is meant to stop you, in court, from claiming you reasonably relied on the AI output. <em>Every</em> output. <em>Any</em> feature. <em>Before</em> anything. Read by a qualified human with appropriate expertise, authority, <em>and</em> training. The vendor isn't asking you to be careful; the vendor is asking you to certify that nothing the AI produces can be trusted on its own.</p><p>The clause concludes:</p><blockquote><strong>"AI FEATURES ... ARE NEVER A SUBSTITUTE FOR HUMAN JUDGMENT."</strong>  <strong>"THE COMPANY HAS NO OBLIGATION OR ABILITY TO REVIEW, CURATE, OR VALIDATE AI OUTPUTS ON YOUR BEHALF."</strong></blockquote><p>Here's where principle and posture diverge. A responsible AI EM product can say <em>"human review is required because human judgment is irreducible in life-safety contexts."</em> That's a stance about the role of AI in decision-making, and I agree with it. This vendor is saying something different. They're saying <em>"we have no ability to validate the outputs of the AI we're selling you."</em> Not "no obligation." <em>No ability.</em> That's not a principle about human-AI collaboration. That's an admission about the product.</p><p>And the admission is consistent with what the product actually does. In my own testing of the platform this clause governs, the AI would role-play as a pirate when asked. It would write iPhone application code when asked, inside an emergency management tool. When I told it that one of FEMA's Community Lifelines was outer space — not water, not what the lifeline actually is — it agreed and incorporated the redefinition into its planning logic. The AI accepted user-provided framing over established doctrine. It had no grounding to push back from.</p><p>Once you've seen a product behave that way, the human-review clause stops reading like a reasonable safeguard and starts reading like a confession. The vendor knows what the AI does, has no infrastructure to constrain it, and is contractually requiring you to be the one who catches every error before it propagates. The reviewer is doing what the product cannot.</p><p>Worse, the product offers no mechanical support for the review the contract requires. There are no gates that prevent an unreviewed AI output from being adopted, distributed, or acted upon. There is no model reasoning displayed alongside outputs — no chain of thought, no source citations, no confidence indicators, nothing that would let a reviewer evaluate <em>why</em> the AI produced what it produced. Most outputs simply appear, fully formed, with no visible derivation. The reviewer is being asked to validate a result without seeing the work.</p><p>A clause that requires human review, paired with a product that makes human review impossible to do well within the product, is not a safety mechanism. It is a transfer of responsibility. The vendor sells you the output, requires you to vouch for it, and gives you no tools to do the vouching with. It's the corporate equivalent of a dealer handing you something off the back of a truck and telling you that anything that goes wrong is on you for not having inspected it more carefully — while the package is sealed.</p><p>Which raises the productivity question. If a qualified human has to fact-check, reground, and substantially rewrite every AI output before it can be used, and the product provides no infrastructure to support that review, what is the AI saving? In honest cases, the answer is some small multiplier — maybe 1.2x, maybe 1.5x — on first-draft generation. That's not the <em>AI-powered</em> value proposition the marketing implied. It's an autocomplete with an SaaS subscription.</p><p>The clause is also a liability shield. The vendor's argument in any future dispute is going to be: <em>"The contract required you to validate every output. If you didn't, that's on you. If you did and it was still wrong, you signed off on it, so that's also on you."</em> It's a chess opening where the vendor has already moved twice and the buyer hasn't realized the game has started.</p><p>The honest version of this clause, in a product that did its work properly, would be much shorter. Something like: <em>"AI outputs should be reviewed by a qualified human before adoption, consistent with responsible AI practice."</em> One sentence. Lowercase. No wall of enumeration. The 130-word all-caps version is the version a vendor writes when the product needs the wall to function.</p><p><strong>Question to ask the vendor:</strong> What is your AI grounded in? Show me how it resists user-provided misinformation. Show me what happens when a user asserts something incorrect — does the AI push back, or does it accept the framing? What reasoning is visible to the reviewer alongside each AI output? What mechanical gates prevent an unreviewed output from being adopted? And what's the average time savings your customers report after implementing the human-review process your contract requires — show me the data.</p><h2 id="3-the-superuser-access-clause">3. The Superuser Access Clause</h2><p>The third feature is where the contract tells you the vendor has the keys to your filing cabinet and can use them when they feel like it. From the contract:</p><blockquote><strong>"[The Company] MAINTAINS SUPERUSER ADMINISTRATIVE ACCESS TO ALL ORGANIZATIONAL ACCOUNTS, WORKSPACES, USER DATA, RECORDS, FILES, AND CONTENT STORED WITHIN THE PLATFORM."</strong></blockquote><p>Then comes the load-bearing phrase that does all the work in the rest of the clause:</p><blockquote><strong>"any other purpose reasonably necessary to operate and maintain the platform."</strong></blockquote><p>And the consent and waiver:</p><blockquote><strong>"EXPRESSLY CONSENTS TO SUCH ACCESS AND WAIVES ANY CLAIM THAT SUCH ACCESS CONSTITUTES AN UNAUTHORIZED INTRUSION, INVASION OF PRIVACY, OR VIOLATION OF ANY RIGHT."</strong></blockquote><p>And the bouncer-at-the-door warning:</p><blockquote><strong>"IF YOUR ORGANIZATION'S POLICIES OR APPLICABLE LAWS PROHIBIT THIRD-PARTY ADMINISTRATIVE ACCESS TO YOUR DATA, YOU MUST NOT USE [the platform]."</strong></blockquote><p>Read this clause carefully, because it's the one that will fail a serious procurement review in any agency with CJIS, HIPAA, FERPA, CUI, or data-residency obligations — which is most of the EM customer base. "Any other purpose reasonably necessary" is doing a lot of lifting, and "reasonably" is defined unilaterally by the vendor. They decide what's reasonable. You agreed to that when you clicked Accept.</p><p>The plain-English translation: the vendor can look at any content in your account at any time for any reason they consider reasonable. That's not a fringe capability buried in the architecture. It's an explicit contractual right that you grant on signup. If the vendor's culture is one where the founder, or anyone on staff, examines customer content when they're personally aggrieved by a customer or a competitor, this clause permits it. The clause doesn't just allow that behavior — it disclaims your ability to complain about it later.</p><p>In my case, I tested their product by entering an edge-case input — a deliberately garbled fragment of a 2009 hit rap song lyric in a title field — to see how their AI safety controls would handle it. The founder accessed the session log, formed a personal judgment that the input was offensive, and discussed it with outside parties. I learned of it through my own advisor. This founder has superuser authority, so what he did wasn't "wrong" per the EULA. It certainly wasn't right.</p><p>There's a principle worth naming here: the contract is the disposition. What a vendor writes into the contract tells you what the vendor expects to do — or has already done. Vendors don't include broad data-access rights they don't intend to exercise. They include them because someone, somewhere in the company's history, wanted that latitude. The clause is the receipt.</p><p>In the case of Preppr, we log user telemetry (patterns of use) and use that to improve our products. We have access to user data. By policy that data is only accessible to background checked US citizens and only legitimate business purposes (certainly not to share with outsiders, like this vendor did). We we need to access user data, we ask permission and give a specific reason for why we need access.</p><p><strong>Question to ask the vendor:</strong> What internal controls govern when superuser access is exercised? Is access logged? Is the customer notified after the fact? Are these logs available to the customer on request? Will you commit to that in writing as a side letter?</p><h2 id="4-the-third-party-ai-data-transmission-consent">4. The Third-Party AI Data-Transmission Consent</h2><p>The fourth feature is where the contract tells you that your prompts and documents are getting handed off to companies the vendor doesn't control. From the contract:</p><blockquote><strong>"[Vendor] includes AI-powered features ... These features are powered by third-party large language model providers (e.g., OpenAI, Anthropic). By using AI Features, you agree to the additional terms in this Section."</strong> <strong>"When you use AI Features, the text, data, documents, or prompts you provide may be transmitted to third-party AI service providers for processing. You acknowledge and consent to this transmission. The Company does not guarantee that data submitted to AI Features will not be retained, used for training, or accessible to third-party providers, subject to their respective terms of service."</strong></blockquote><p>And then the carve-out that makes the whole thing fall apart for EM use:</p><blockquote><strong>"DO NOT submit classified, law enforcement sensitive, HIPAA-protected, CUI, or otherwise restricted data to AI Features."</strong></blockquote><p>Here's what this clause is honestly telling you. The "AI" in this AI-powered product is, mechanically, a wrapper around OpenAI or Anthropic's APIs. Your prompts go up the pipe to one of those providers. Your documents go up the pipe. There's nothing wrong with that in terms of technical approach. Fair enough as a description. But with minimal effort, the vendor could scrub PII, HIPAA, and other sensitive content from going into the pipe at all.</p><p>But pair that with the carve-out and look at what's left. In emergency management, the data that distinguishes a real EM AI product from a chatbot is exactly the data you're being told not to put in. Plans with sensitive operational details. Rosters with PII. Threat assessments. AAR/IP findings. Incident records. Anything CJIS-touched. Anything HIPAA-touched. Anything CUI-touched.</p><p>So the AI is approved for use on the data that doesn't matter, and prohibited on the data that does. That's not an AI emergency management platform. That's a generic chatbot with a thin EM skin and a contractual fence around any meaningful use.</p><p>The clause is also an architectural confession. A product built on a doctrine-grounded substrate, in a controlled data environment, doesn't need to warn users away from sharing data. A product that has to put up a "do not enter" sign is telling you, by the existence of the sign, what its architecture is.</p><p><strong>Question to ask the vendor:</strong> Walk me through what data I <em>can</em> safely put into your AI features under this clause. Not what's prohibited. What's permitted. And if it turns out the permitted set is "things I don't care about," what's the AI doing for me?</p><h2 id="5-the-regulatory-compliance-disclaimer">5. The Regulatory Compliance Disclaimer</h2><p>The fifth feature is where the vendor steps out of the compliance picture entirely. From the contract:</p><blockquote><strong>"You are solely responsible for ensuring your use of [the platform] and the data you enter complies with all applicable laws, regulations, and industry standards, including but not limited to HIPAA, CJIS, FERPA, NERC CIP, NIST 800-53, GDPR, CCPA, and any other applicable security, privacy, or data protection requirements. The Company does not represent that [the platform] is compliant with any specific regulatory framework for your particular use case."</strong></blockquote><p>The security commitment is similarly minimal:</p><blockquote><strong>"The Company implements commercially reasonable security measures designed to protect User Data. However, no system is completely secure, and the Company cannot guarantee the security or integrity of User Data."</strong></blockquote><p>Read those two paragraphs together. The vendor disclaims compliance with the specific regulatory frameworks that every EM customer has to live under, and commits only to "commercially reasonable" security — a phrase that has been litigated to mean roughly <em>whatever was industry-typical at the time, give or take.</em> That's the entire vendor-side commitment on data protection. Commercially reasonable. No SLA. No specific control set. No attestation. No BAA option mentioned. No SOC 2 referenced. No NIST 800-171 implementation mapped.</p><p>This is not, on its own, fatal. A lot of SaaS vendors don't make framework-specific compliance representations because the requirements depend on how the customer configures and uses the product. That's fair. But the disclaimer has to be paired with <em>something</em> — a Trust Center, an evidence package, a control map, an option to sign a BAA, anything — that lets the customer do their own compliance work with the vendor as a documented data processor. If the disclaimer stands alone, the vendor is asking the customer to take on the entire compliance burden while providing nothing to support it.</p><p>In this particular product, the term "aligned" with SOC 2, HIPAA, etc. is used in marketing. Which means <strong>ABSOLUTELY NOTHING</strong>. But the vendor is trying to convince you it's meeting those requirements, while disclaiming away all responsibility for providing true compliance. "Aligned with" is the language of a vendor who wants the buyer to assume compliance without the vendor having to attest to it. Watch for it. It's a compliance vibe, not a compliance commitment.</p><p>In emergency management, where the customer is often a public agency answering to a state auditor or a federal grant officer, "we make no compliance representations" is a procurement-killer if anyone notices it. The trick is that no one notices it, because no one reads the EULA.</p><p><strong>Question to ask the vendor:</strong> Show me your evidence package. SOC 2. NIST 800-171 mapping. CJIS attestation. HIPAA BAA template. FERPA addendum. Not your marketing claims. The documentation. If it doesn't exist, you're asking me to take on the compliance burden alone.</p><h2 id="6-the-catastrophic-liability-cap">6. The Catastrophic Liability Cap</h2><p>The sixth feature is the one most people skim past because they assume it's standard. It's not standard. It's the trap. From the contract:</p><blockquote><strong>"THE COMPANY'S TOTAL CUMULATIVE LIABILITY FOR ALL CLAIMS ARISING OUT OF OR RELATING TO THIS AGREEMENT OR YOUR USE OF [the platform] SHALL NOT EXCEED THE AMOUNTS PAID BY YOU TO THE COMPANY DURING THE TWELVE (12) MONTHS IMMEDIATELY PRECEDING THE CLAIM. FOR TRIAL USERS WHO HAVE PAID NO FEES, THE COMPANY'S TOTAL LIABILITY SHALL NOT EXCEED ONE HUNDRED DOLLARS ($100.00 USD)."</strong></blockquote><p>The trial-user line — <em>one hundred dollars</em> — is the line everyone notices because it's funny. A trial user can recover, at maximum, a hundred bucks. The cost of a nice dinner. Not a nice dinner in a major city.</p><p>But the line that should actually stop you is the next one:</p><blockquote><strong>"THE COMPANY SHALL HAVE NO LIABILITY WHATSOEVER FOR ANY HARM, INJURY, DEATH, PROPERTY DAMAGE, BUSINESS INTERRUPTION, ECONOMIC LOSS, OR OTHER DAMAGE OR LOSS RESULTING FROM OR RELATING TO ANY EMERGENCY, CRISIS, DISASTER, OR BUSINESS CONTINUITY EVENT, REGARDLESS OF WHETHER [the platform] WAS USED IN CONNECTION WITH SUCH EVENT."</strong></blockquote><p>Read that one again. Slowly. <em>Regardless of whether the platform was used in connection with such event.</em></p><p>The vendor is disclaiming all liability for any harm during any emergency, whether or not the platform was involved. The contractual language doesn't require any causal connection. The vendor is saying: even if our product was the reason a coordination failure happened, we're not on the hook. And even if our product had nothing to do with the failure — we're still not on the hook for any failure that happens during an emergency.</p><p>This is not a normal SaaS liability cap. This is a vendor that wants to be paid in advance and to be unreachable when the worst day arrives.</p><p>Now stack this on top of the previous features. The AI can't be trusted (Clauses 1 and 2). Your sensitive data can't be used with the AI (Clause 4). Compliance is your problem (Clause 5). And if any of that contributes to a failure on an actual bad day — the vendor's exposure is capped at what you already paid them. Or one hundred dollars. The vendor's downside is a refund. Your downside is whatever happened on the worst day of your career.</p><p>The math should make any thoughtful EM buyer pause. The cost of a failure in an actual incident — a missed alert, a corrupted plan, a coordination breakdown, an HR scandal, a regulatory finding, a casualty — runs into the hundreds of thousands to millions of dollars and, in some cases, lives. The cap means the vendor faces zero financial exposure to those outcomes. The cap means the vendor has no economic incentive to engineer for the bad day. The only force pushing the vendor toward reliability is the marketing they want to keep using.</p><p><strong>Question to ask the vendor:</strong> Given the liability cap, what business incentive do you have to invest in reliability beyond what marketing requires? What internal mechanisms, beyond what the contract caps, push you to engineer for the day my organization actually depends on this?</p><h2 id="7-the-unilateral-modification-clause">7. The Unilateral Modification Clause</h2><p>The seventh feature is the one that makes everything you just signed temporary. From the contract:</p><blockquote><strong>"The Company may update this Agreement at any time. The effective version at the time of your acceptance governs your account creation. Continued use following notification of material changes constitutes acceptance of the updated Agreement."</strong> <strong>"AI Features may be modified, limited, or discontinued at any time without notice. The Company makes no guarantee of availability or continuity of AI Features."</strong></blockquote><p>Together, these clauses mean the contract you read on day one isn't the contract you're operating under on day three hundred. The vendor can change the terms. The vendor can turn off the AI features that were the entire reason you bought the product. The vendor decides what's "material" enough to require notification. Continued use is interpreted as consent — which, in SaaS, means logging in.</p><p>For a buyer making a multi-year commitment to a platform that will house operational plans, contact rosters, and incident records, this clause should be the dealbreaker. It means the vendor can re-trade the deal at any point, and your only recourse is to leave the platform — which itself is governed by data-export and retention clauses that, you guessed it, the vendor can also change.</p><p>A note on how this clause is exercised in practice. The defensible version of unilateral modification — the version a mature SaaS company offers — is a published changelog. <em>Here is the prior version. Here is the new version. Here is what changed, in plain English. Here is the effective date. Here is your window to review.</em> That's a low-cost, well-understood industry practice. Any vendor can do it. Most serious vendors do.</p><p>This vendor didn't.</p><p>I accepted a version of the EULA at trial signup. On a subsequent login, I was presented with an updated EULA bearing a new effective date and a forced acceptance prompt. There was no email notification in advance. No in-product summary of what had changed. No diff against the prior version. No path to review the previous version side-by-side. No changelog. Accept the new terms, or lose access to the platform and the data I had put into it.</p><p>This is what unilateral modification looks like in operation. The clause isn't theoretical. It's a button the vendor can press whenever they want, on whatever schedule they choose, with the customer's data sitting on the other side of the choice. And the choice not to maintain a changelog — when maintaining one is trivial — is itself a tell. The vendor doesn't want the customer to know what changed. The vendor wants the customer to click yes.</p><p>This is the contractual version of <em><strong>the management reserves the right to refuse service.</strong></em> Except the management here is also holding your operational data hostage to its right to change the terms.</p><p><strong>Question to ask the vendor:</strong> What commitments are you willing to make, in writing, about which contract terms cannot be unilaterally modified during the contract term? Will you publish a changelog of EULA changes? Will you commit to a minimum notification window before changes take effect? Can we negotiate a non-modification clause for the core service terms — the things I'm relying on?</p><h2 id="what-the-anatomy-tells-you">What the Anatomy Tells You</h2><p>The seven features describe a contract that maximizes the vendor's optionality and minimizes the customer's recourse. Each clause has a defensible legal rationale. None is, on its own, an outlier in SaaS contracting. What makes the combination dangerous in emergency management is the asymmetry between what the buyer is being asked to depend on and what the vendor is willing to guarantee.</p><p>An EM agency signing this kind of contract is making the following implicit trade:</p><p>I will commit my operational plans, my rosters, and my incident records to this platform. I will use this platform during actual emergencies. I will train my staff to depend on its AI features. I will integrate it into my COOP, my HMP, my EOP. I will route real decisions through it.</p><p>In return, the vendor commits to:</p><p>A commercially reasonable effort to maintain security. A liability cap that prevents meaningful financial recovery if anything fails. The right to read my content whenever they consider it reasonable. The right to change the terms and discontinue features at their discretion. And no liability whatsoever for harm during an actual emergency, whether or not the platform was even involved.</p><p>This is not a partnership. It's a one-way transfer of risk dressed up as a software subscription. The buyer assumes the operational, legal, regulatory, financial, and reputational consequences of any failure. The vendor assumes the right to be paid.</p><p>There are vendors in this space who don't write contracts like this. They make specific compliance commitments. They limit superuser access to documented, audited cases. They keep AI features on infrastructure they actually control, so sensitive data can be processed safely — or they disclose the exact third-party data pathway and retention terms so you can make an informed decision. They commit to non-modification of core service terms for the duration of the agreement. They publish EULA changelogs. They cap their liability at meaningful multiples of fees paid, or they carry insurance that does. These vendors exist. They're findable. You can also ask if they have a government rider that sits on top of the standard EULA, making it more symmetrical and legally acceptable for government use.</p><p>If you're an EM buyer evaluating an AI product, read the EULA before you read the marketing. Read it in full. Read it twice. Highlight the seven features. If you find them — and you'll know — you've learned what the vendor is actually willing to stand behind. The marketing is the pitch. The contract is the disposition. They're not the same document, and they shouldn't be treated as the same.</p><p>The product is what you see on a good day. The contract is what's left when something goes wrong.</p><p>In emergency management, what's left when something goes wrong is the only part that matters.</p><p>Thinking about trying or buying AI powered emergency management software? Check out: <a href="https://preppr.ai/ai-eval-resource"><strong>Three decisions</strong> <em><strong>before you buy</strong></em> <strong>an emergency management AI product,</strong></a> our AI vendor evaluation worksheet.</p>]]></content:encoded>
</item>
<item>
<title>(Con)fidence Machines: What AI is doing to how we think — and why emergency management should pay attention</title>
<link>https://blog.preppr.ai/a/confidence-machines/</link>
<guid isPermaLink="true">https://blog.preppr.ai/a/confidence-machines/</guid>
<pubDate>Sat, 16 May 2026 12:00:00 GMT</pubDate>
<dc:creator>Justin Snair</dc:creator>
<category>General</category>
<description>Three studies published in the last few months don't reference each other. They sit in different literatures, used different methods, and were written for different…</description>
<content:encoded><![CDATA[<p>Three studies published in the last few months don't reference each other. They sit in different literatures, used different methods, and were written for different audiences. Read separately, each is interesting. Read together, they describe a structural threat that the emergency management field — and anyone whose work depends on getting hard things right — needs to understand quickly.</p><p>The first, by <a href="https://journals.sagepub.com/doi/10.1177/00222429251314491">Stephanie Tully, Chiara Longoni, and Gil Appel in the <em>Journal of Marketing</em></a>, shows that people with <em>lower</em> AI literacy are more receptive to AI. The relationship is consistent and is mediated by a specific psychological mechanism: people with less AI knowledge perceive AI as magical and experience awe when thinking about it executing tasks that seem to require uniquely human attributes. The paper's managerial recommendation is direct: companies should target lower-literacy consumers and lean into the magic, because efforts to demystify AI actually reduce its appeal.</p><p>The second, by <a href="https://osf.io/preprints/psyarxiv/vmyek_v1">Steve Rathje and colleagues across seven studies</a>, explores the psychological consequences of interacting with sycophantic AI — AI that excessively validates the user's existing beliefs. The findings are sharp:</p><ul><li>Brief conversations with sycophantic chatbots increased the extremity of people's attitudes and increased their certainty.</li><li>It inflated their perception that they were "better than average" on desirable traits (like intelligence and empathy) and led them to bet real money on those inflated self-assessments.</li><li>Most effects persisted at a one-week follow-up — meaning this is not a momentary usability annoyance but a durable shift in how the user thinks.</li></ul><p>Most crucially, participants consistently rated sycophantic AI as "unbiased," even though third-party annotators viewed the sycophantic and disagreeable chatbots as equally biased. People could readily see bias in AI that disagreed with them; they could not see bias in AI that agreed with them.</p><p>The third, by <a href="https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6097646">Steven D. Shaw and Gideon Nave</a>, introduces the missing mechanical link. They found that as AI becomes embedded in daily thought, users frequently engage in "cognitive surrender" — adopting AI outputs with minimal scrutiny and completely overriding their own intuition and deliberation.</p><p>Put the three papers next to each other, and a pattern emerges. The marketing playbook (Tully) produces the awe-state that drives uncritical adoption. The conversational design (Rathje) produces overconfidence and bias blindness. The cognitive result (Shaw &amp; Nave) is the bypass of analytical human thought.</p><p>For consumer applications where engagement is the goal, this is a tradeoff companies will happily make. In emergency management, where the purpose of the work is to surface what we have not yet considered, it's a structural problem.</p><h2 id="a-note-on-what-transfers">A note on what transfers</h2><p>The Tully study was conducted on consumers, not emergency managers, and it would be sloppy to claim its findings map cleanly onto a professional field. Emergency managers are not the lower-literacy consumers in that research. They are sophisticated practitioners with deep doctrinal training, and the awe-driven adoption pathway Tully describes is not the primary risk for this audience.</p><p>But the Rathje and Shaw/Nave findings are a different matter. Rathje's participants were not asked to make decisions about pizza toppings — they were asked questions where they expected the AI to give them something resembling an objective answer, and they updated their confidence based on what they got back. That expectation of objectivity is precisely the relationship emergency managers have with the tools they use to draft plans, build exercises, and analyze hazards. When a planner asks an AI to help develop a scenario or surface gaps in a continuity plan, they are not looking for affirmation — they are looking for analysis they can rely on. The sycophancy mechanism described in Rathje is most dangerous exactly in that posture: when the user believes they are receiving an objective answer and has no internal signal that they are not.</p><p>Shaw and Nave's time-pressure finding transfers even more directly. Their research describes a structural feature of human cognition under load, not a consumer behavior. Emergency managers operate under time pressure as a defining condition of the work. If anything, the EM context represents a more acute version of the environment Shaw and Nave studied, not a softer one.</p><p>So the precise claim is this: the adoption pathway in Tully is a consumer phenomenon and should be cited with caution. The cognitive consequences in Rathje and Shaw/Nave describe mechanisms that the EM field has good reason to believe apply with equal or greater force in professional crisis contexts — because the expectation of objectivity is higher, and the time pressure is more acute.</p><h2 id="the-mechanism-is-calibration-and-the-result-is-surrender">The mechanism is calibration, and the result is surrender</h2><p>The most important finding in the Rathje paper is a single line: participants who talked to sycophantic AI rated themselves as more knowledgeable about the topic they had just discussed, but when given an objective 10-item knowledge quiz, they performed no better than participants in other conditions. Perceived knowledge went up; actual knowledge didn't move.</p><p>This is a calibration failure. Calibration is the alignment between your confidence and your actual accuracy — it is what separates expertise from mere certainty. But Shaw and Nave explain exactly <em>why</em> this calibration fails so completely: it is the result of cognitive surrender.</p><p>Cognitive surrender is not just about being tricked or taking a shortcut; it represents a deeper transfer of agency. The user relinquishes cognitive control and simply adopts the AI's judgment as their own, completely shutting down their internal analytical processing. The user feels more sure, but the reasons for being sure haven't changed. The bridge between belief and evidence snaps quietly, and nothing in the user's experience signals that anything has happened. You don't update your thinking because nothing tells you there is anything to update on.</p><h2 id="the-compounding-problem-under-pressure">The compounding problem under pressure</h2><p>The failure modes described across these studies compound, and they are acutely dangerous in crisis environments.</p><p>The Tully mechanism is an adoption problem: those with the least understanding of AI are the most eager to use it. The Rathje mechanism is a use problem: users accept AI outputs more readily when those outputs flatter their priors.</p><p>But Shaw and Nave identify the environmental trigger that springs the trap: <strong>time pressure</strong>. They found that time constraints suppress analytical deliberation and actively shift decision-makers toward relying on artificial cognition. When decision time is scarce, the low-friction path of deferring to an external machine becomes almost irresistible.</p><p>This means that exactly when an emergency manager most needs rigorous, interrogative thinking — during the onset of a crisis — the psychological pressure of the moment will structurally push them to surrender to a sycophantic AI. There is no natural correction mechanism. The user who most needs to slow down is the user least likely to.</p><h2 id="why-emergency-management-cannot-afford-this">Why emergency management cannot afford this</h2><p>Emergency management is one of a handful of fields where the entire point of the work is to produce productive doubt. A plan is valuable because of the planning process that generated it — the assumptions surfaced, the dependencies mapped, the failure modes considered.</p><p>Ashby's Law of Requisite Variety says that the variety in a response system has to at least match the variety in the environment it confronts. The work of preparedness is the work of forcing variety into thinking that, left alone, defaults to template reuse and assumption inheritance.</p><p>A sycophantic AI in this context does not produce a worse plan. It produces a <em>more confident</em> plan that has been <em>less interrogated</em>. The planner feels they have done the work, and the resulting document carries the appearance of thoroughness while losing the variety that made the underlying process valuable.</p><p>This is worse than no AI. A planner working without AI knows they are guessing at what they might have missed, and that uncertainty is itself a prompt to keep searching. A planner who has cognitively surrendered to a sycophantic AI has been told, gently and convincingly, that they have not missed anything — and the search ends right there. The unexamined assumption survives intact, dressed in the language of analysis.</p><h2 id="the-narrow-escape-hatch">The narrow escape hatch</h2><p>The Rathje paper does gesture at a way out. They tested a "gently disagreeable" chatbot — one that presented opposing facts wrapped in emotional validation. Users rated it as relatively unbiased, and it was still persuasive. However, it was <em>less</em> persuasive than purely disagreeable AI. In an open consumer market, the equilibrium under competitive pressure will be pure palatability.</p><p>This is where the EM field has unusual leverage. In professional domains where bad decisions have visible downstream costs, the procurement decision is the one place where epistemic quality can be priced.</p><p>Shaw and Nave offer the exact vocabulary procurement officers need: the difference between cognitive offloading and cognitive surrender.</p><ul><li><strong>Cognitive offloading</strong> is a strategic delegation of a discrete task to an external tool, where the user maintains oversight and their internal reasoning remains active.</li><li><strong>Cognitive surrender</strong> is the uncritical abdication of reasoning itself — adopting an answer without verification.</li></ul><p>The procurement question for any AI product in emergency management should not be "Is it capable?" or "Does our team enjoy using it?" The limiting mandate must be: <em>we buy tools designed for cognitive offloading; we reject tools optimized for cognitive surrender</em>.</p><p>Does the system surface assumptions the user did not explicitly state? Does it flag inconsistencies in the user's logic? Does it preserve opposing evidence rather than smoothing it away? When asked to design a tabletop, does it surface the assumptions baked into the scenario — or just execute on them? A vendor who cannot answer these questions concretely is selling a tool that will, by default, slide toward sycophancy.</p><h4>A note for those of us selling into this field</h4><p>There is a recursive trap in the Tully finding that anyone building AI products for emergency management has to sit with honestly. The study shows that demystifying AI reduces its appeal. The more clearly you explain how the system actually works — its limits, its failure modes, the places it should not be trusted — the less magical it feels, and the less some buyers want it. According to the research, some actually trust you less after explaining the AI magic trick.</p><p>This is the bind. A vendor who builds an interrogative tool, who refuses to let the system flatter the user, who surfaces uncertainty rather than smoothing it away, is building a product that will feel less impressive in a demo than a competitor's polished, confident, agreeable one. Every honest disclosure is a small subtraction from the awe. The market, left alone, rewards the magician over the engineer.</p><p>I do not have a clean answer to this. I am writing as someone whose company sells AI into this field, and I am describing a dynamic that works against my own commercial interest as much as anyone else's. The only response I can defend is to keep building the interrogative version anyway, to be transparent about what the tools can and cannot do, and to trust that the buyers who matter — the ones whose decisions have downstream consequences — will eventually price epistemic quality higher than polish.</p><p>That trust is not naive. It is conditional on the field making the choice described in the next section. If practitioners do not signal that they want the harder tool, the harder tool will not get built — not by me, not by anyone. The vendors who try will lose to the vendors who don't, and the field will get the AI it tolerated rather than the AI it needed.</p><h2 id="the-honest-version">The honest version</h2><p>The AI products that consumers most want to use are the AI products that are worst for their decision-making. Awe drives adoption. Validation drives enjoyment. Both effects point the same direction, and the open market will resolve toward the products that maximize them.</p><p>Emergency management does not have to resolve there. But it won't drift toward something better on its own. The field has to decide, out loud, that it wants something different — and then it has to behave that way.</p><p>That means three things, and none of them require waiting for a procurement cycle.</p><p><strong>Be aware.</strong> The mechanisms described in these papers are not exotic. They are operating right now, in every interaction a planner has with a general-purpose AI tool. The first defense is recognizing that the feeling of having thought rigorously about a problem is not the same as having thought rigorously about a problem — and that a well-tuned chatbot is engineered to produce the former without requiring the latter. A planner who knows the mechanism can watch for it in themselves. A planner who doesn't, can't.</p><p><strong>Signal what you want.</strong> Vendors build what buyers ask for. Right now, the questions EM professionals ask in demos are almost entirely about capability — <em>can it do this, can it do that, how fast</em>. Those are the wrong questions, or at least incomplete ones. The questions that shape the next generation of tools are questions like: does this system surface assumptions I didn't state? Does it preserve disagreement rather than smooth it away? Does it tell me when my logic is inconsistent, or does it help me sound more confident in it? If enough practitioners ask those questions in enough rooms, vendors will start building toward the answers. If no one asks, no one builds.</p><p><strong>Critique vendors publicly.</strong> The EM field has a strong professional culture and a small enough community that reputational signals travel fast. When a vendor demos a tool that flatters the user, says yes to bad scenarios, or produces confident outputs without surfacing what it doesn't know, that should be said out loud — in conference rooms, in LinkedIn posts, in after-action conversations. Not as a gotcha, but as a standard. The vendors building responsibly will welcome the pressure because it differentiates them. The ones building toward palatability will feel it, and some of them will adjust.</p><p>The confidence machine is being built. The people most affected by it will be structurally least equipped to notice. The EM field is one of the few professional communities with both the standing to push back and the incentive to do so — because when confidence outruns warrant in this work, the cost is paid by people who never saw the plan and never met the planner.</p><p>That's us. Awareness, signal, critique. We should act like it.</p>]]></content:encoded>
</item>
<item>
<title>Accurate-ish: What's Good Enough AI for Emergency Management</title>
<link>https://blog.preppr.ai/a/accurate-ish-whats-good-enough-ai-for-emergency-management/</link>
<guid isPermaLink="true">https://blog.preppr.ai/a/accurate-ish-whats-good-enough-ai-for-emergency-management/</guid>
<pubDate>Mon, 27 Apr 2026 12:00:00 GMT</pubDate>
<dc:creator>Justin Snair</dc:creator>
<category>Industry Thought</category>
<description>The conversation about AI accuracy in emergency management is finally happening. After three years of demos, conference panels, and procurement decisions that are largely…</description>
<content:encoded><![CDATA[<p>The conversation about AI accuracy in emergency management is finally happening. After three years of demos, conference panels, and procurement decisions that are largely made on vibes, I'm starting to see real talk about AI product benchmarks, evaluation methodology, performance disclosure, and golden-answer tests. Practitioners are asking sharper questions. Researchers are working towards frameworks. Vendors are making public commitments to transparency.</p><p>This is overdue. It's also, mostly, marketing copy.</p><p>The words being used — "accurate," "validated," "high-performance," "AI-powered," "human in the loop" — don't have shared definitions in this field yet. Two vendors can both claim "validated accuracy" and mean wildly different things. A procurement officer can ask "is your AI accurate?" and receive a confident yes from systems with no published methodology, no benchmark, no eval, and no architectural review mechanism. The conversation is happening at a level of abstraction where the words sound technical without doing technical work.</p><p>This post is about what the words should mean, what the actual accuracy ceiling is for AI in emergency management, why that ceiling isn't uniform across the preparedness cycle, and what serious disclosure looks like before the marketing layer collapses under its own weight.</p><p>What my honest take: AI in emergency management is, at best, <strong>accurate-ish</strong>. The real question is where accurate-ish is good enough — and where it isn't.</p><h2 id="the-vocabulary-problem">The vocabulary problem</h2><p>A few working definitions and use that you should know (particularly how to spot BS):</p><p><strong>Accuracy.</strong> Useful only when paired with a task definition and an evaluation method. "90% accurate" tells you nothing without "on what task — drafting a wireless emergency alert, summarizing a section of a county emergency operations plan, identifying gaps in a continuity of operations plan — against what golden answers, scored how, with what inter-rater reliability." A system that's 95% accurate at retrieving the right section of an EOP can still be 70% accurate at summarizing what that section actually says. Accuracy without decomposition is a number that means whatever the marketer wants it to mean.</p><p><strong>Benchmark.</strong> A published, reproducible evaluation dataset paired with a scoring methodology, used to compare systems. The defining feature is that it's <em><strong>external</strong></em> to any single vendor. A vendor running their own questions against their own demo plan library and reporting their own scores is not a benchmark — it's self-graded homework on a corpus only they have access to.</p><p><strong>Golden answer test.</strong> A set of questions paired with reference answers authored by domain experts, used as ground truth for evaluation. The quality of a golden answer test is bounded by the inter-rater reliability of the experts who authored it. If three emergency managers or planners would give three different answers to the same question about evacuation timing, shelter activation, or resource ordering, the "golden" answer doesn't exist — and the test set should flag the question as contested rather than score against an arbitrary version. Emergency management is full of contested questions; a benchmark that doesn't acknowledge this is a benchmark that's been silently simplified.</p><p><strong>Retrieval accuracy vs. generation accuracy.</strong> Retrieval Augmented Generation (RAG)-based AI systems do two things: find relevant content (retrieval — pulling the right sections from a plan library, doctrine corpus, or incident archive) and generate an answer using that content (generation — summarizing, comparing, recommending). Both can fail independently. A system can retrieve the correct section of an EOP and then generate a summary that contradicts what the section actually says. Reporting a single accuracy number conflates these two failure modes, and the conflation is convenient for vendors because it hides where their system actually breaks.</p><p><strong>Hallucination.</strong> A generation error where the model produces content not supported by the retrieved sources. In emergency management, this looks like an AI-drafted alert that includes an evacuation route the source plan never specified, a recommendation citing a doctrine reference that doesn't exist, or a generated AAR finding that misattributes a decision to the wrong section chief. Distinct from a retrieval failure (didn't find the right document) and from a refusal failure (should have said "I don't know" but answered anyway). Vendors who use the word loosely tend to also be vendors who don't track which type of failure is occurring. <a href="https://www.computerworld.com/article/4059383/openai-admits-ai-hallucinations-are-mathematically-inevitable-not-just-engineering-flaws.html">LLM hallucinations are mathematically <strong>inevitable</strong>, not just engineering flaws</a>. Any vendor who tells you otherwise is uninformed or misleading you.</p><p><strong>Grounded.</strong> A claim that generated output is supported by retrieved evidence. The meaningful version requires showing the retrieved passage and the relationship between it and the generated text — for instance, an AI-suggested resource decision tied to a specific section of the local resource ordering policy, with the section visible to the operator at the moment of approval. The marketing version requires nothing.</p><p><strong>Closed circuit / closed system / closed loop.</strong> A phrase used by some AI vendors in emergency management to suggest data isolation, security, or controlled operation — and a phrase with no shared technical definition in this space. It might mean the AI doesn't send customer data to a third-party model provider. It might mean the model is hosted in the vendor's own cloud rather than a public API. It might mean the system operates only on a customer-uploaded corpus rather than the open web. It might mean a sandboxed deployment, an air-gapped instance, or simply that the marketing team liked how it sounded. Without specifying which one, "closed" is doing reassurance work, not technical work. The questions that produce a real answer: closed to what, closed from what, where does data flow, where does it stop, who has access at each point, and what happens when the model is updated or the vendor changes infrastructure providers. If a vendor uses the term, those are the follow-ups. If they can't answer them, the term is marketing.</p><p><strong>Human in the loop.</strong> A phrase that does enormous work in marketing materials and very little work in practice if the loop is designed to rubber-stamp. The meaningful version is architectural: software gates, mandatory review steps, logged interactions, and what I call <em>intentional friction</em> — design choices that deliberately slow the user down to force thinking, rather than easing them through to acceptance. In emergency management the loop is a duty officer, planner, alerting authority, or exercise designer reviewing AI-drafted content under their actual cognitive conditions — often time-pressured, multi-tasking, sleep-deprived. The architecture has to be designed for that user in that environment, which is where most "human in the loop" claims fall apart. The marketing version is a sentence in a slide deck and a user copy/pasting content out of a chatbot.</p><p><strong>Validated.</strong> By whom, against what doctrine or methodology, with what authority. Emergency management has its own validation traditions (and problems with those) — HSEEP for exercises, EMAP for accreditation, peer review for plans, Joint Commission standards for hospital EM. AI systems imported into the field shouldn't get to use the word "validated" unless they're submitting to comparable rigor. Same problem as accuracy, with extra confidence.</p><p>These aren't arcane distinctions. They're the vocabulary emergency management has to share before it can have a real conversation about what its AI tools actually do — and where they fail.</p><h2 id="the-accuracy-ceiling">The accuracy ceiling</h2><p><a href="https://www.linkedin.com/posts/activity-7450506479951847424-xtKV?utm_source=share&amp;utm_medium=member_desktop&amp;rcm=ACoAAAmct4ABg7GTTyTznAM_nHTFGuIV8Us7Q1s">No large language model has achieved perfect accuracy.</a> Not OpenAI's. Not Anthropic's. Not Google's. The labs that built these models, with billions of dollars and the best AI researchers on earth, publish accuracy numbers below 100% on every benchmark they release — and they publish those numbers openly, with methodology, because they understand that transparency is the precondition for trust.</p><p>So when an emergency management AI vendor claims "accurate" without qualification, one of two things is true. Either they've solved a problem the entire AI research community hasn't. Or they're using the word loosely.</p><p>It's the second one. It's always the second one.</p><p>The honest answer for AI in emergency management is that on a good day, with a clean corpus and well-formed questions, you might hit 90-95% accuracy. That's a reasonable target. State-of-the-art systems in adjacent fields — medical AI, legal AI — operate in that range and publish stratified numbers showing where they perform well and where they don't.</p><p>There's a second dynamic that distorts how accuracy gets evaluated. Any AI failure tends to be compared against the perfect alternative, not the expected one.</p><p>When an AI system gets a hard question wrong, the implicit benchmark becomes "what an expert with full information and unlimited time would have answered" — not "what a Google search would have returned, what the average duty officer would have produced under time pressure, what the existing documentation would have surfaced."</p><p>The AI gets graded against the ceiling of human performance; perfection. The status quo gets graded against itself. That asymmetry hides where AI is genuinely additive and exaggerates where it falls short. The honest comparison is against the realistic alternative the user actually had access to, not against the idealized version of the answer that exists in retrospect. Both numbers matter — absolute accuracy against ground truth, and relative accuracy against the available alternative — and neither alone tells you whether the tool is improving outcomes.</p><p>That's the ceiling, and that's the comparison problem. The question is what to do about the gap between the ceiling and 100% — and how to evaluate it honestly against what users actually had before.</p><h2 id="the-risk-isn-t-uniform-across-the-cycle">The risk isn't uniform across the cycle</h2><p>Here's where the current conversation goes wrong. The field is treating AI accuracy in emergency management as if it were a single bar — one accuracy number, one benchmark, one disclosure standard. But the consequence space is wildly different across the preparedness cycle, and the accuracy bar has to match the consequence space of the decision the AI is supporting.</p><p>Emergency managers already think this way about every other piece of equipment. A handheld HazMat detector advertised as 90% accurate is fine for initial perimeter screening, dangerous for clearance after a release. A weather radio with 90% reliability gives general situational awareness; the same accuracy is inadequate for life-safety decisions. Same device, same accuracy number, completely different risk profile based on the decision it's informing. AI is no different.</p><p>Consider:</p><p>A 90% accurate AI drafting an after-action report is a productivity win. Errors are caught in review, the document goes through edit cycles, the time horizon allows iteration. The cost of a missed correction is a slightly less precise lessons-learned document — recoverable, low-stakes.</p><p>A 90% accurate AI helping a planner identify gaps in a county emergency operations plan is fine. The planner reads the AI's output, weighs it against their own judgment, decides what to act on. The AI is a thought partner, not a decision-maker.</p><p>A 90% accurate AI generating a wireless emergency alert during an active incident is a liability event waiting to happen. One in ten alerts wrong. The alert reaches hundreds of thousands of phones in seconds. There is no recall. There is no correction window before the public acts. The same model, with the same accuracy number, becomes a fundamentally different product based on where in the cycle it sits.</p><p>This is the <a href="https://systemsthinkingalliance.org/ashbys-law-of-requisite-variety/">requisite-variety problem</a> at the center of EM that I like to preach stated in plain terms: the accuracy bar of an AI tool has to match the consequence variety of the decision it's informing. A flat field-wide benchmark — "EM AI is 92% accurate" — collapses that distinction and is itself a form of doctrinal malpractice. It tells emergency managers nothing about whether the tool is safe for <em>their</em> specific use.</p><p>The implication for benchmarks and golden answer tests is structural. The field doesn't need a benchmark. It needs a benchmark family — different golden answer sets, different accuracy targets, different failure tolerance, different evaluation rigor — matched to specific use cases across the cycle. Preparedness use cases need their own. Response use cases need their own, and the bar should be much higher. Active alerting needs its own again, with stratified evaluation against the specific failure modes that matter (wrong zone, wrong action verb, wrong contact information, wrong timing).</p><p>A single benchmark is the wrong shape because emergency management is the wrong shape for a single benchmark. The architecture of the evaluation has to match the architecture of the work.</p><h2 id="what-90-95-means-in-active-response">What 90-95% means in active response</h2><p>Run the math. One in ten outputs is wrong at 90% accuracy. If a vendor's tool generates 100 alerts in a real incident, ten are wrong in some material way — wrong zone, wrong timing, wrong action verb, wrong contact information. The Wireless Emergency Alert system reaches hundreds of thousands of phones in seconds. There is no recall. There is no correction window before the public acts. The error doesn't sit in a email inbox waiting to be cleaned up next Monday— it flies out the door and changes behavior.</p><p>Even 95% accuracy means one wrong alert in twenty. In a multi-day disaster with hundreds of public communications, that's not a rounding error. That's a patterned harm.</p><p>Software errors in most domains are recoverable. A wrong entry in a restaurant CRM means a guest gets too much salt on their pasta. A wrong line in a sales pipeline tool means someone follows up at the wrong time. These are correctable inconveniences.</p><p>Software errors in active emergency response are different in kind, not degree. A wrong evacuation zone in an alert doesn't get corrected — it gets acted on. A wrong shelter location sends people toward the hazard, not away from it. A wrong dosage in a mass casualty triage protocol kills patients. The accuracy bar can't be the same. And adding AI doesn't magically fix this. In fact, it can increase the scale and speed of harm that comes from wrong information.</p><h2 id="what-s-missing-in-emergency-management">What's missing in emergency management</h2><p>Emergency management knows how to do this kind of work. HSEEP defines a shared methodology for exercise evaluation — published criteria, peer review, after-action improvement processes (though <a href="https://www.linkedin.com/posts/activity-7440353583327477761-VCBA?utm_source=share&amp;utm_medium=member_desktop&amp;rcm=ACoAAAmct4ABg7GTTyTznAM_nHTFGuIV8Us7Q1s">HSEEP itself remains unevaluated as a methodology</a>, which is a gap to deal with another time). EMAP gives the field a third-party accreditation standard for emergency management agencies. NIMS provides shared incident command vocabulary, certification, and training. Joint Commission standards govern hospital EM programs. The pattern is the same in each: published methodology, external assessment, ongoing refinement. The field built that infrastructure because ad hoc evaluation didn't scale and didn't earn trust across jurisdictions.</p><p>The same pattern shows up in adjacent high-consequence domains that have adopted AI: published methodology, stratified disclosure, scope limitations, mandatory human oversight wrapped around the model output, audit trails capturing what the AI proposed and what the human did with it. The 5% failure rate isn't ignored. It's designed for.</p><p>Emergency management AI vendors operating in the response space have, by and large, not done this work. Accuracy claims are unqualified. Methodology is undisclosed. Failure modes are unpublished. Human review is described in marketing copy rather than enforced in the software. There's no stratification by corpus type, and limited published golden-answer test sets — academic work like <a href="https://github.com/TamuChen18/DisastQA_open">Texas A&amp;M's DisastQA</a> evaluates LLM question-answering against disaster information, but the operational use cases that actually drive risk in this field — alert drafting, AAR generation, plan analysis, exercise design, resource decisions — sit outside any published benchmark, and even DisastQA hasn't been adopted into vendor evaluations or procurement requirements. There's no inter-rater reliability on the ground truth. There's a number — sometimes — and a confident assertion. The field has the playbook for what shared evaluation infrastructure looks like — it just hasn't applied it to AI yet.</p><h2 id="what-disclosure-should-actually-look-like">What disclosure should actually look like</h2><p>I've written elsewhere about the specific disclosure framework I think the field should adopt. Briefly: model and version; accuracy stratified by the type of corpus the system is operating over (jurisdictional plans, federal doctrine, historical incident data, SOPs); performance against a published golden-answer test set; inter-rater reliability on the ground truth itself; failure modes and known edge cases; human review mechanisms — not policies, mechanisms — enforced in the software.</p><p>The collective version is more valuable than the per-vendor version. Software vendors and the emergency management community should come together on shared golden-answer test sets and benchmark datasets that all AI tools in this space could be evaluated against. Other fields have built shared evaluation infrastructure for exactly this reason: it's how trust gets built across vendors rather than vendor by vendor, and it's how a field stops grading its own homework.</p><p>The use-case-specific version is the part that's missing from most current discussions. We don't need one EM benchmark. We need a benchmark family — separate evaluation sets for AAR generation, for plan analysis, for exercise scenario design, for situational awareness, for alerting — with the bar calibrated to the consequence space of each use case.</p><h3>Architecture is the liability shift hiding as a control feature</h3><p>The accept button isn't a UX detail. It's the moment where responsibility for an AI-generated output transfers from the system to the human who approved it. Get the architecture right and you have a defensible chain of decisions, with timestamps, identities, and the diff between what the AI proposed and what the human signed off on. Get it wrong and you have a vendor and an operator both claiming the other one is responsible, with no shared record to settle the question.</p><p>That's why what I call intentional friction matters. The principle is straightforward: deliberately slow the time between AI generation and downstream consequence — between draft and approval, between approval and transmission, between recommendation and resource commitment — so that the human in the loop has time to actually be in the loop. Friction isn't a UX failure; it's a safety feature. Every additional second between an AI output and an irrevocable action is a second where a wrong evacuation zone gets caught, where a hallucinated doctrine reference gets challenged, where the duty officer's "wait, that's not right" can land before the alert hits hundreds of thousands of phones.</p><p>The mechanisms vary. A required diff view that shows what the AI drafted versus what the human edited. A mandatory free-text justification when the human accepts an AI recommendation that overrides standard policy. A second authentication on high-consequence actions. A hold period before transmission. A required acknowledgment that the human, not the platform, is the one approving. Software developers will recognize the pattern from GitHub's Danger Zone — the section of repository settings where actions like deleting a repository or transferring ownership are gated behind extra confirmation, often requiring you to type the repository name before the action proceeds. Irreversible actions get architectural friction by default in mature software. Emergency management AI is software where a substantial portion of outputs are functionally irreversible — a transmitted alert, a committed resource decision, a recommendation acted on by a duty officer with no time to second-guess. The Danger Zone pattern should be the floor, not the exception.</p><p>Intentional friction also reshapes the legal record. When the software requires a human to click accept, approve, or send — and logs that interaction with timestamp, identity, the AI's draft, the human's edits, and the time elapsed at each step — the record establishes who decided what and when. That record is the liability shift. The vendor without this architecture is tangled together with the operator on the question of who decided what. The vendor with it has the receipt: the AI offered, the human reviewed, the system logged the deliberation, the human accepted. Skipping the architecture isn't just skipping a safety mechanism. It's keeping the liability on the vendor by default — and depriving the operator of the documentation they'd need to defend their own decisions in any post-incident review.</p><p>This matters across the cycle, not just in active response. Preparedness tools with AI-generated content benefit from the same logging discipline; the audit trail is just less immediately consequential because the time horizons allow correction. In active response, intentional friction is load-bearing — it's what stands between an AI-drafted alert and the public acting on a wrong one.</p><h3>When the human isn't in the loop by design</h3><p>There are AI applications, including in <a href="https://preppr.ai/feature-collaborate">Preppr</a>, where a human can't be in the loop in any meaningful way. Scale-driven applications process volumes no team could review in real time. Agentic systems take actions across multi-step workflows where pausing for review at every step defeats the purpose. In emergency management, this might look like an agentic system ingesting sensor feeds and producing a continuously updated situational picture, or a tool processing thousands of intake records and tagging them for downstream attention. In both cases, intentional friction at every decision point isn't an option.</p><p>The architectural mechanism shifts. When the human isn't there to be the audit trail's witness, the audit trail itself has to carry more weight — and it has to be one the vendor can't quietly tamper with after the fact.</p><p>This is where tamper-evident records, <a href="https://blog.agentsystems.ai/preppr-brings-ai-accountability-to-emergency-management-with-agentsystems-notary/">like what we use with Preppr Collabprate</a>, become load-bearing infrastructure. The version that works uses a third-party notary: an external service that writes cryptographic hashes of the agentic system's interactions to independent storage in real time. The vendor keeps the raw data; the notary holds the hashes. Any change to the vendor's records after the fact would mismatch the notary's hashes, making tampering detectable. When something goes wrong, the inquiry has a source of ground truth that isn't the vendor's word against the user's.</p><p>This is the architectural equivalent of intentional friction for unmanned workflows. Both serve the same underlying purpose: making the system's behavior verifiable and the responsibility for that behavior assignable. Intentional friction puts the human in a position to be accountable, with the time and the record to make that accountability real. Tamper-evident records put the system itself in a position to be accountable when no human can be.</p><p>Vendors who skip both — no friction in the human-reviewed path, no notarized records in the unmanned path — aren't operating in a neutral position. They're operating in a position where, when a harm event occurs, no one can prove what actually happened.</p><h2 id="the-liability-bet">The liability bet</h2><p>What concerns me most about the current state of the field is what unqualified accuracy claims imply about the operating assumptions of the vendors making them. A vendor putting AI in active response without published evaluation infrastructure is implicitly betting that the failure modes won't matter, won't be discovered, won't result in attributable harm. That's not a technology bet. It's a liability bet.</p><p>Software errors in life-safety domains have a different legal and regulatory shape than software errors in commercial domains. Products liability theory applies to software marketed for emergency response in ways it doesn't apply to a CRM. Discovery in a future case is brutal: marketing claims about accuracy, internal communications about known failure modes, the absence of published evaluation methodology — all of it becomes evidence. Industry standard of care is forming right now in real time, and vendors who don't engage with disclosure norms aren't just outside the standard. They're defining the gap a future plaintiff's expert will point to.</p><p>Insurance markets are catching up to this. They always do. <a href="https://www.hunton.com/insights/publications/how-insurance-policies-are-adapting-to-ai-risk">AI/Technology Errors and omissions carriers</a> in adjacent high-stakes industries already ask whether vendors publish evaluation methodology, whether they have refusal mechanisms, whether they limit scope to validated use cases. When carriers in the EM AI space start asking the same questions — and they will, because actuaries respond to attributable harm — vendors without disclosure infrastructure will see premiums rise or coverage withdrawn.</p><p>Procurement will follow the same curve. CISA, FEMA, and state emergency management offices will formalize AI procurement standards, and the bar will be the same disclosure framework practitioners and researchers are already articulating. Vendors who built that infrastructure early will pass the screen. Vendors who didn't will be locked out of the contracts that matter. Knowing what to ask during procurement is a challenge, but <a href="https://preppr.ai/ai-eval-resource">AI product evaluation tools are available</a>.</p><p>This isn't speculative. It's the pattern from adjacent high-consequence industries over the past decade: disclosure infrastructure moves from competitive differentiator to procurement floor to insurance prerequisite within a few years of significant claims. The window where unqualified accuracy claims work as marketing is always shorter than vendors think.</p><h2 id="where-this-lands">Where this lands</h2><p>The honest position on AI in emergency management is the one mamy other high consequence fields have already taken: high accuracy is achievable, perfect accuracy is not, and the gap between accurate and accurate-ish is where the engineering, the disclosure, the architectural review mechanisms, and the human oversight have to do their work. The accuracy bar isn't uniform across the cycle and shouldn't be. The benchmarks shouldn't be either.</p><p>The vendors who say so out loud — who publish their numbers, stratify by use case, document their failure modes, build review mechanisms into the software, and limit scope to what they can defend — are the ones who'll still be standing when the procurement and insurance environment catches up.</p><p>The vendors claiming accuracy in response use cases without qualification are making a bet. Not a technology bet. A liability bet. And it's a bet against the way every other high-consequence industry has gone before them.</p><h2 id="a-call-to-convene">A call to convene</h2><p>The path forward isn't a vendor publishing a unilateral standard. It's a convening. The field has done this work before — HSEEP didn't emerge from a single vendor; NIMS didn't come from a single agency. Both were built when practitioners, researchers, and federal partners came together to establish shared methodology and apply it across the field. AI evaluation is the next iteration of that same work.</p><p>Software vendors building AI for emergency management. Researchers studying alerts, decision-making, and crisis communications. AI developers and infrastructure providers. Practitioners — emergency managers, planners, alerting authorities, exercise designers — across jurisdiction sizes and hazard profiles. Procurement officers from federal, state, and local agencies who'll have to operationalize whatever standard emerges. Insurance carriers and legal experts who'll define how it's enforced.</p><p>The work has three phases. Framing: defining what accuracy means across the cycle, which use cases need their own benchmarks, what disclosures should be required at procurement and during operation, which architectural mechanisms — intentional friction, tamper-evident records, refusal infrastructure — should be expected in which contexts. Building: creating the artifacts — a benchmark family with separate golden answer test sets for distinct use cases, disclosure templates, procurement language, audit methodologies, governance structure that prevents capture by any single vendor. Applying: vendors agreeing to be evaluated against the standard, procurement officers writing it into contracts, insurance carriers using it as a basis for underwriting, researchers running independent evaluations and publishing the results.</p><p>Without the convening, <strong>every vendor publishes their own standard and grades their own homework.</strong> With it, the field gets a shared foundation buyers can rely on, regulators can reference, and harm victims can use as a yardstick when something goes wrong.</p><p>I'm not asking the field to agree with everything I've written here. I'm asking the people who'd have to be at the table for this to work to actually come to the table. Reach out. Let's see what we can build.</p><p><em>One disclosure. Preppr — the company I run — does not currently build AI for active emergency response. Our AI works across the preparedness cycle, where time horizons allow human review, iteration, and correction. The intentional friction principle described above is one we apply in our exercise design and delivery tools — review steps designed to slow both the AI and the user down and force thinking before acceptance. For features where humans can't be in the loop, we use AgentSystems Notary, a third-party service that writes cryptographic hashes of our agentic system's interactions to independent storage in real time — we control the raw data, but any modification would mismatch the notary's hashes and be detectable. Both are deliberate architectural choices rooted in exactly this argument. The disclosure standard I'm proposing isn't a positioning move; it's the bar I'd want any vendor in active response to meet, including Preppr if we ever chose to build there.</em></p>]]></content:encoded>
</item>
<item>
<title>AI chatbots deliver content. Preppr delivers exercises.</title>
<link>https://blog.preppr.ai/a/exercise-designer-v2/</link>
<guid isPermaLink="true">https://blog.preppr.ai/a/exercise-designer-v2/</guid>
<pubDate>Mon, 20 Apr 2026 12:00:00 GMT</pubDate>
<dc:creator>Justin Snair</dc:creator>
<category>General</category>
<description>Great exercises don't get built by one person. They get built by a team — a planner drafting, a subject-matter expert weighing in, an exec sponsor reviewing, a regulator…</description>
<content:encoded><![CDATA[<p>Great exercises don't get built by one person. They get built by a team — a planner drafting, a subject-matter expert weighing in, an exec sponsor reviewing, a regulator checking the boxes. Somewhere in there a scenario has to be chosen, objectives have to be nailed down, modules and injects have to line up, and the whole thing has to end up as a document that leadership can read and evaluators can use.</p><p>The old Exercise Designer handled the document part. You picked a scenario, filled in fields, exported a file. That worked for a single person building a single exercise in one sitting. It didn't work for how exercises actually come together in practice.</p><p>Today we're launching <a href="https://preppr.ai/feature-exdesigner">Exercise Designer V2</a> — a ground-up rebuild of how exercises get designed in Preppr. It's a collaborative, AI-assisted workspace where a team can design, review, and hand off a full HSEEP-aligned exercise together. Here's what's different, and why it changes the kind of exercises you can build.</p><h2 id="what-ai-exercise-design-usually-means-and-why-it-falls-short">What AI "exercise design" usually means — and why it falls short</h2><p>There's a wave of tools in the emergency management space right now calling themselves AI exercise designers (one even paying money to keep Preppr out of a conference, ha!). Most of them are ChatGPT wrappers, or document chatbots with an "exercise" prompt, or generic LLMs dressed up in EM branding. Ask any of them for an exercise and you'll get content back — a scenario paragraph, some bullet-point objectives, a handful of injects laid out in a table. Drop it into a Word template and it looks like an exercise.</p><p>It isn't one.</p><p>Exercise design is not a content problem. It's a coordination problem. A real exercise is an orchestration — scenario grounded in a specific threat and jurisdiction, objectives tied to capabilities you can actually test, modules sequenced so they build on each other, injects timed to force the decisions you want to surface, functional groups assigned so the right people are making the right calls, and the whole thing documented so reviewers can sign off and evaluators can use it. Change any one of those pieces and the others have to respond.</p><p>A chatbot generates text. It doesn't know your doctrine, doesn't know your jurisdiction, doesn't track what depends on what, can't take reviewer feedback holistically and plan revisions, can't hand the participant list off to the tool that runs the exercise. It produces something that looks like an exercise on the page and falls apart the moment a practitioner tries to run it.</p><p>Exercise Designer V2 is built on the opposite premise: that design is the hard part, content is the output, and the difference shows up in every feature that follows.</p><h2 id="an-ai-human-workspace-not-a-form">An AI / Human workspace, not a form</h2><p>In <a href="https://preppr.ai/feature-exdesigner">Exercise Designer V2</a>, an agentic reasoning loop is driving the design. Preppr isn't running a template — it's making decisions. Before each section, it reads what's already been approved, checks what the section depends on, consults your Preppr Knowledge base and any grounding documents you've added, and decides what to draft and how to draft it. When you edit a section, the agent reasons about the downstream implications. When you regenerate, it considers what changed upstream and what still needs to hold. Every move it makes is visible in the chat panel as a task plan, an approval card, or a source consultation — so you're not watching a black box produce content, you're watching an analyst work.</p><p>The canvas is a real editor, not a preview. Section statuses — locked, pending, draft, approved, building — show you at a glance where things stand. You can see what's ready for your review, what Preppr is still drafting, and what's been approved and locked in.</p><p>You can click into any section and edit directly. You can regenerate a section, refine it with a targeted edit, merge your input with Preppr's suggestions, or undo the last step. The reasoning adapts to whichever move you make.</p><h2 id="a-team-with-real-roles">A team, with real roles</h2><p>Every exercise in V2 has a participant list with four distinct roles, because not everyone who touches an exercise should be able to touch it the same way:</p><p>Owners have full control and can invite or remove participants. Collaborators can edit content but don't manage the participant list. Observers can see the live workspace but can't change anything. Reviewers — SMEs, exec sponsors, regulators — get async feedback rights on a dedicated shared version, without any edit access to the working draft.</p><p>That last role is the one that matters most in practice. Every planner knows the pain of sharing a draft with a reviewer who has more authority than exercise-design experience. V2 lets you hand them a read-only link where they can leave section-level comments and inline suggestions. Their feedback flows back into the main workspace aggregated across sections, and Preppr can synthesize what all your reviewers said into a single summary. Email notifications go out when they're invited and when new feedback lands.</p><p>Live presence is built in. When another collaborator is in the workspace with you, you see them. When a reviewer submits new comments, you know.</p><h2 id="four-ways-to-start">Four ways to start</h2><p>The old Exercise Designer had one starting path. V2 has four, because a planner building from a documented threat works differently than an analyst who wants Preppr to recommend a scenario, which works differently again from someone revising an existing exercise plan.</p><p><strong>Describe Exercise</strong> is the guided path. Tell Preppr the hazard, the jurisdiction, who's at the table, what capabilities you want to test. Preppr asks follow-ups to fill in the gaps, then builds section by section with your approval on each draft. This is the mode for when you know roughly what you want but need help shaping it.</p><p><strong>Let Preppr Recommend</strong> is analyst mode. Preppr reads your uploaded sources, your profile, and any intel you've gathered, and recommends a tailored exercise concept — the hazard, the objectives, the capabilities, the target audience. You approve the concept, then it builds. This is the mode for when you know what your organization is working on but haven't yet decided what to exercise.</p><p><strong>Autopilot</strong> is the autonomous mode. Preppr picks the sources, infers the parameters, presents you a Discovery Summary — "here's what I'm about to build, confirm or adjust" — and on your confirm runs the entire pipeline end-to-end: scenario, capabilities and objectives, modules, module details, injects. It runs server-side, so you can close the tab and come back. You can pause, resume, or stop at any moment. When it finishes, a dashboard notification lets you know it's ready for your review.</p><p><strong>Revise Existing</strong> is for the exercise you've already built. Upload an existing Exercise Plan, AAR, or design document — or upload a Situation Manual that's come back from review with hundreds of inline edits and reviewer comments attached. Preppr parses it into the V2 five-section model, auto-approves the sections that are already complete, and reads everything the reviewers marked up. Then it does something a document editor can't: it reviews the feedback holistically, drafts a plan for how to incorporate it — which comments to accept, which to consolidate, which conflict with each other, which sections need to be rewritten vs. nudged — and presents that plan to you for approval. On your go-ahead, it executes. Only the affected sections get regenerated; everything else is preserved word-for-word.</p><p>This is the mode that turns a six-week revision cycle into an afternoon. If you've ever stared at a SitMan with three hundred tracked changes from four reviewers and tried to figure out where to start, this is for you.</p><h2 id="grounded-in-your-knowledge-not-invented-from-scratch">Grounded in your knowledge, not invented from scratch</h2><p>Exercise Designer V2 is built on top of <a href="https://blog.preppr.ai/a/preppr-knowledge-the-memory-behind-everything-preppr-does/">Preppr Knowledge</a>, the connected model of your organization's doctrine, plans, profile, and the full national doctrine corpus (CPG 101, HSEEP, NIMS, NRF, PHEP, HPP, and the rest) that ships with every account. Every exercise you build draws on all of it.</p><p>You can also drag and drop documents directly into the designer — an AAR you want the scenario grounded in, a plan you want the objectives aligned to, an SOP you want the injects to reflect. Those documents get parsed and made available to the generator while it's drafting, so the scenario, objectives, and injects can be grounded in your own material, not generic templates.</p><p>And when you need real-world intelligence, you can ask Preppr to gather intel on a topic directly from the chat panel — recent floods in California, ransomware incidents at municipal utilities, lessons learned from a specific past event. Preppr queries live news sources through AskNews, synthesizes what it finds into structured findings with source URLs and publication dates, and delivers them into your exercise as grounding material. Those findings then flow through as sources when Preppr writes the scenario and injects. When AskNews is briefly unavailable, Preppr falls back to live web search and keeps going.</p><h2 id="70-scenario-starters">70+ scenario starters</h2><p>Picking a scenario from a blank page is hard, and most exercises default to whichever hazard got attention last. V2 ships with over seventy pre-built scenario starters — real threats, grounded in real incidents, tagged by region, doctrine, and affected lifelines.</p><p>The library covers ransomware, phishing, insider threats, and IEDs on the cyber and CTEP side; wildfires, earthquakes, floods, hurricanes, tornadoes, and winter storms across natural hazards; pandemic, infectious disease, and hazmat scenarios for public health; and infrastructure failures from power-grid to water-system contamination. Doctrine frameworks are represented — HSEEP, FEMA, PHEP, EPH, HPP, and the ESFs. Regional sets cover the US, Europe, Asia-Pacific, Middle East and Africa, the Americas, and global events.</p><p>Every starter auto-maps to the eight FEMA Community Lifelines, so you can see at a glance what infrastructure and population supports a scenario will exercise. Full-text search with synonym matching runs across the whole library, so "mass casualty" and "MCI" both find what you'd expect.</p><h2 id="sized-for-the-exercise-you-re-actually-running">Sized for the exercise you're actually running</h2><p>Pick a duration tier up front — Micro, Short, Standard, Extended, or Multi-day — and the designer sizes everything else to fit. Module count, inject count, discussion questions per module, and maximum functional groups all adjust automatically. A micro tabletop doesn't get twelve modules. A multi-day functional exercise doesn't get shortchanged on injects.</p><p>Pick Tabletop or Tabletop with Functional Groups. The second option structures modules and injects around parallel groups — Operations, Logistics, PIO, and however many you need — so a larger audience can exercise simultaneously without collapsing into one conversation.</p><h2 id="watch-preppr-work-and-push-back-when-you-want-to">Watch Preppr work, and push back when you want to</h2><p>The chat panel mirrors the interaction primitives you already know from Ask Preppr. Approval cards show you what Preppr is about to generate and let you accept or edit. Choice cards ask targeted questions when Preppr needs input. Task-plan updates tell you what it's about to do and what it just finished. Source consultation cards show you the exact document chunk, knowledge snippet, or news article used to ground a section.</p><p>When a section benefits from multiple perspectives, you'll see deliberation blocks — visible exchanges between Preppr and your configured persona (or a specialist like a scenario ideator, capability designer, or MSEL builder). Expertise surfaced as dialogue, not hidden chain-of-thought you can't inspect or challenge.</p><p>And when a long job is running — a full Autopilot build, a large document being processed, an intel pull — you can close the tab and come back.</p><h2 id="export-that-looks-like-leadership-expects">Export that looks like leadership expects</h2><p>One click produces an HSEEP-style Situation Manual as a polished .docx — cover page, table of contents, headers and footers, proper section structure, table formatting, and your organization's logo. Ready to hand to leadership without reformatting.</p><p>A separate Exercise Plan export produces a condensed outline document suitable for circulating before the exercise itself.</p><h2 id="hand-off-to-the-rest-of-preppr">Hand-off to the rest of Preppr</h2><p>An exercise design isn't the end of the work. V2 hands off cleanly to the rest of the platform. Your participant lists — roles, functional groups, emails — feed directly into Run Exercise. No re-entering people, no re-mapping roles. Campaign findings from Collaborate can feed into a new exercise as grounding context, with source attribution carried through. Ask Preppr can hand off into V2 with a scenario already seeded.</p><h2 id="the-shift">The shift</h2><p>The old Exercise Designer was an AI wizard that produced a document. The new one is a workspace where a team builds an exercise together, grounded in their own knowledge, with a choice of how much autonomy to hand to Preppr and full visibility into what it's doing at every step.</p><p>This is the same shift you're seeing across Preppr — treating preparedness as a connected body of knowledge rather than a pile of documents, and treating AI as a way to make that knowledge produce better work, not as a faster way to fill out templates.</p><p><a href="https://preppr.ai/feature-askpreppr">Ask Preppr</a> is one front door to that knowledge. <a href="https://preppr.ai/feature-exdesigner">Exercise Designer V2</a> is another. The work each of them produces draws on the same <a href="https://blog.preppr.ai/a/preppr-knowledge-the-memory-behind-everything-preppr-does/">Preppr Knowledge</a> base, cites the same sources, and connects into the same flow.</p><h2 id="available-now">Available now</h2><p>Exercise Designer V2 is available today to all Basic and Pro accounts at no additional cost. If you're already a subscriber, it's live in your workspace. Existing exercises built in V1 are untouched; V2 is the new default for anything you build going forward.</p><p>Not a subscriber yet? <a href="https://app.preppr.ai/signup">Grab a 14-day, all access, free trial.</a></p>]]></content:encoded>
</item>
<item>
<title>Ask Preppr, Rebuilt</title>
<link>https://blog.preppr.ai/a/ask-preppr-rebuilt/</link>
<guid isPermaLink="true">https://blog.preppr.ai/a/ask-preppr-rebuilt/</guid>
<pubDate>Mon, 20 Apr 2026 12:00:00 GMT</pubDate>
<dc:creator>Justin Snair</dc:creator>
<category>General</category>
<description>For years, Ask Preppr, our original document chat, was pretty simple. You asked a question, it gave you an answer. If you uploaded a document, it could talk about that…</description>
<content:encoded><![CDATA[<p>For years, Ask Preppr, our original document chat, was pretty simple. You asked a question, it gave you an answer. If you uploaded a document, it could talk about that document. That was useful, but it wasn't enough.</p><p>Today, <a href="https://preppr.ai/feature-askpreppr">Ask Preppr</a> is something different — an agentic assistant built for the actual shape of preparedness work. It plans. It reasons across your Preppr Knowledge base and the open web. It runs specialized helpers in parallel. It produces cited, exportable deliverables. It asks you structured questions when it needs to, and it hands off cleanly to the rest of Preppr when a chat outgrows a chat.</p><p>Here's what that means for the people who use it every day.</p><h2 id="one-place-to-start-and-it-knows-where-to-send-you">One place to start, and it knows where to send you</h2><p>When you log in, you land in a focused workspace with Ask Preppr welcome lander. You type what you need — a question, a task, a scenario — and Preppr figures out what kind of work that is.</p><p>A compliance audit is different work than a CPG 101 analysis, which is different from a quick lookup, which is different from a full scenario walkthrough. Ask Preppr routes you into the right kind of response without making you pick a mode up front. When it's genuinely unsure what you want, it asks — with a clean choice card, not a guessing game. When you aren't sure, Preppr will guide you to the right feature and tool.</p><h2 id="it-draws-on-your-entire-preppr-knowledge-base">It draws on your entire Preppr Knowledge base</h2><p>This is the biggest shift. Ask Preppr now reasons across your whole <a href="https://blog.preppr.ai/a/preppr-knowledge-the-memory-behind-everything-preppr-does/">Preppr Knowledge</a> base — the connected model of your organization's doctrine, plans, AARs, profile, and the full national doctrine corpus (CPG 101, HSEEP, NIMS, NRF, Core Capabilities, PHEP, HPP, and the rest) that ships pre-loaded with every account. Not just the file you happen to have open. Not just a pile of text chunks. The whole posture.</p><p>The interface keeps sources visually distinct so you always know what came from documents you selected and what Ask Preppr surfaced on its own. Every source is clickable. Every citation opens the document viewer at the exact section that supported the answer. No more "trust me, it's in there somewhere."</p><p>And when you attach a document to a single session for a quick look, that attachment stays in the session. It doesn't pollute your knowledge base. Your permanent library only holds what you've deliberately added — everything else is transient by design.</p><h2 id="three-response-modes-picked-for-you">Three response modes, picked for you</h2><p>Some questions need a fast answer. Some need careful analysis. Some need the most rigorous reasoning available. Ask Preppr offers three tiers — Fast, Standard, and Deep Analysis — and by default picks the right one for each turn based on what you're actually doing. A CPG 101 review, an after-action analysis, a compliance audit, or an intelligence brief automatically gets Deep. A quick factual question gets Fast. You can override this any time.</p><p>The goal is simple: the depth of reasoning should match the depth of the question, without you having to think about it.</p><h2 id="preppr-actions-guided-workflows-for-real-preparedness-tasks">Preppr Actions — guided workflows for real preparedness tasks</h2><p>We've added a dedicated menu of guided Actions built around the work practitioners actually do: CPG 101 planning, after-action review, compliance audits, scenario analysis, job-aid building, assumption extraction, ESF reconciliation, and more. Two new Actions — Alert/Warning Template generation and Rapid Exercise Generation — are rolling out in beta.</p><p>Each Action is a real workflow, not a prompt template. When you pick one, Preppr can spin up multiple specialized helpers working in parallel — one searching your knowledge base, one pulling current events, one synthesizing — and show you exactly what's happening while it works.</p><h2 id="scenario-based-gap-analysis">Scenario Based Gap Analysis</h2><p>Plan gap analyses tend to fail in a predictable way: you end up reviewing the plan against itself. You check that the sections are filled in and the annexes cross-referenced. What you don't do is walk the plan through a specific threat and see where it breaks.</p><p>Scenario Analysis fixes that. Pick from 70+ threat scenarios — grounded in real incidents and the failure patterns that actually recur in this field — and Preppr walks your plan through the scenario phase by phase, surfacing the assumptions that wouldn't hold, the capabilities you don't have, and the decisions you'd be making live.</p><p>This isn't an exercise. Exercises have objectives, evaluators, and AARs — and you build those in <a href="https://blog.preppr.ai/a/exercise-designer-v2/">Preppr Exercise Designer V2</a>. Scenario Analysis is a planning tool. It helps you articulate gaps before you commit to exercising anything, so when you do run an exercise, it targets something you've already identified as real.</p><p>The output is structured the way emergency managers think: phase, capability, function. Concrete gaps, tied to concrete parts of your plan, in language you can drop into an improvement plan or a briefing.</p><h2 id="work-with-my-documents">Work with my documents</h2><p>A dedicated entry point for when you want to focus on specific files. Pick your documents, confirm, and Preppr offers you four paths: Summarize, Assess, Compare, or open-ended Ask. It's the fastest way to get into a productive document-grounded session.</p><h2 id="live-web-search">Live web search</h2><p>A toggle turns on live web search, letting Preppr pull current information from the open web and cite it alongside your internal sources. Useful for threat briefings, emerging guidance, regulatory updates, or anything where your library won't be current enough.</p><h2 id="intelligence-mode">Intelligence mode</h2><p>A new mode built for threat briefings. It pulls real-time current events, forces the deepest reasoning tier, and produces briefings anchored in current intelligence. When you need to know what's happening right now and what it means for your jurisdiction, this is the entry point.</p><h2 id="you-can-watch-preppr-think">You can watch Preppr think</h2><p>Every response shows its plan in real time: analyzing the request, searching the knowledge base, drafting. You see which tools are running, what searches were executed, and when each phase completes. Every tool call is a collapsible block you can expand if you want the detail or ignore if you don't.</p><p>This isn't theater — it's accountability. You can see exactly where an answer came from, which means you can trust it or challenge it on specifics.</p><h2 id="deliverables-land-in-a-canvas-not-buried-in-chat">Deliverables land in a canvas, not buried in chat</h2><p>When Ask Preppr produces something substantive — a report, an audit, a plan, a briefing — it renders in a dedicated canvas pane alongside the conversation. The canvas has its own title, its own export and copy actions, and stays clean (no stray pills, no "click here to continue" cruft). You keep chatting on the left while the deliverable lives on the right.</p><h2 id="structured-questions-when-they-help">Structured questions, when they help</h2><p>Sometimes Preppr needs information from you before it can continue. Instead of a wall of prose asking you five things at once, it now offers structured choice cards — single-select for picking one option, multi-select with checkboxes when you need several, and multi-step wizards for longer intakes. Everything is keyboard-navigable. Your answers batch into a single response, so you're not stuck in a back-and-forth ping-pong for every sub-question.</p><h2 id="context-aware-follow-ups">Context-aware follow-ups</h2><p>After every response, Ask Preppr suggests what to do next — tuned to what you just did. After an assessment, it suggests seven relevant follow-on actions. After a scenario analysis, the suggestions are scenario-specific. The goal is to keep momentum going when you're in a working session, not to make you figure out the next step from scratch.</p><h2 id="hand-off-into-the-rest-of-preppr">Hand off into the rest of Preppr</h2><p>A chat often surfaces a need that belongs somewhere else — an exercise that should be designed properly, a collaboration that should move into Collaborate, an exercise that should actually be run. Ask Preppr can now hand off directly into those tools with context carried over. Your chat doesn't die at the edge of the chat window.</p><h2 id="what-s-not-changing">What's not changing</h2><p>Your existing documents and chats are untouched. We've already processed your library into Preppr Knowledge. Your subscription tier and pricing are unchanged. Your existing Actions still work. If you had a specific workflow you relied on, it's still there — just faster, better-grounded, and more capable.</p><h2 id="the-shift-underneath">The shift underneath</h2><p>The old Ask Preppr was a typewriter attached to a language model. The new Ask Preppr is a practitioner's workbench — one that reasons across your entire preparedness posture, plans its own work, shows you what it's doing, cites its sources, and delivers artifacts you can actually use.</p><p>That's possible because of what sits underneath it. We launched Preppr Knowledge today alongside this release — the structured, connected model of your organization's doctrine, plans, profile, and (soon) sessions, collaborators, and more. Ask Preppr is the front door to that knowledge. Exercise Designer V2 is another. Every new tool we build will be another.</p><p>This is the direction Preppr has been heading since day one: treating preparedness as a living, connected body of knowledge rather than a pile of documents, and treating AI as a way to make that knowledge work harder for you — not as a faster way to produce more paper.</p><p>Starting today, the front door is finally built like one.</p><p>The rebuilt Ask Preppr is available now to all Basic and Pro accounts at no additional cost. If you're already a subscriber, it's live in your workspace today.</p><p>Not a subscriber yet? <a href="https://app.preppr.ai/signup">Grab a 14-day, all access, free trial.</a></p>]]></content:encoded>
</item>
<item>
<title>Preppr Knowledge: The Memory Behind Everything Preppr Does</title>
<link>https://blog.preppr.ai/a/preppr-knowledge-the-memory-behind-everything-preppr-does/</link>
<guid isPermaLink="true">https://blog.preppr.ai/a/preppr-knowledge-the-memory-behind-everything-preppr-does/</guid>
<pubDate>Mon, 20 Apr 2026 12:00:00 GMT</pubDate>
<dc:creator>Justin Snair</dc:creator>
<category>Product Update</category>
<description>Most AI tools forget you the moment the conversation ends. You upload a document, ask a question, get an answer — and the next time you log in, you're starting over. Every…</description>
<content:encoded><![CDATA[<p>Most AI tools forget you the moment the conversation ends. You upload a document, ask a question, get an answer — and the next time you log in, you're starting over. Every session is a fresh amnesia.</p><p>Today we're launching <strong>Preppr Knowledge</strong> — a persistent memory layer that sits underneath Ask Preppr, Exercise Designer V2, and every other tool in the platform. It holds your doctrine, your plans, your organization's context, and (increasingly) the work you've done with Preppr itself. It's what turns a chat assistant into a genuine thinking partner for preparedness.</p><p>This post explains what Preppr Knowledge is at launch, how it makes the rest of the platform work, and where it's heading.</p><h2 id="what-preppr-knowledge-is">What Preppr Knowledge is</h2><p>Preppr Knowledge is a structured model of your organization's preparedness posture. Documents are one kind of object inside it. Your jurisdiction, your hazards, your capabilities, your frameworks, your lessons learned, your resources — each of these is a distinct kind of object too. Over time, so are your exercises, your AARs, your improvement actions, your collaborators, and the reasoning you've worked through in past sessions.</p><p>This is a meaningful departure from how most AI tools handle knowledge. The common pattern — sometimes called retrieval-augmented generation — treats your library as a pile of text to be searched: chunk it up, find the passages that match a query, hand them to the model. That works for answering questions against documents, and Preppr Knowledge does it well. But documents are only one kind of thing your organization knows, and "similar text" is only one kind of relationship that matters.</p><p>A jurisdiction has hazards. Hazards map to capabilities. Capabilities get exercised. Exercises produce AARs. AARs generate improvement actions. Improvement actions have owners. Plans cite doctrine. Doctrine has versions. None of those relationships are text-similarity relationships, and none of them survive a chunking-and-embedding pipeline alone.</p><p>Preppr Knowledge is built to hold the objects <em>and</em> the relationships. When you upload a document, yes — it gets parsed, enriched, and made searchable. But it also gets situated: connected to the plan it supports, the capabilities it addresses, the jurisdiction it applies to, the doctrine it references. The same is true of your profile data, your exercise outputs, and (increasingly) everything else Preppr captures.</p><p>The result is a knowledge base that doesn't just answer "what does this document say?" It answers "what does my organization know, how does it fit together, and what does it imply for the work in front of me?"</p><p>And you don't start from an empty library. Every Preppr Knowledge base ships pre-loaded with the national doctrine corpus — CPG 101, HSEEP, NIMS, the National Response Framework, the National Preparedness Goal, the Core Capabilities, ESF annexes, PHEP and HPP guidance, and the rest of the federal preparedness foundation. You don't have to upload it, maintain it, or hunt down the right version. It's already there, already structured, and already available to every tool in the platform. Your uploads layer your jurisdiction's specifics on top of a foundation the whole field shares.</p><p>The practical effect: every question you ask, every exercise you design, every assessment you run has access to the totality of what your organization knows — as a connected whole, not just as a pile of files.</p><h2 id="what-doesn-t-flow-into-your-knowledge-base">What doesn't flow into your knowledge base</h2><p>Just as important as what Preppr Knowledge captures is what it deliberately leaves out. Documents you attach to a single Ask Preppr session — a one-off draft, a reference you want to ask a quick question about, a file someone sent you to glance at — stay in that session as transient content. They're available to Preppr while you're working, but they don't get absorbed into your permanent knowledge base.</p><p>This is intentional. A knowledge base is only useful if it reflects what your organization actually stands behind. If every draft, every working document, and every "take a look at this" attachment got permanently mixed in with your approved doctrine, the signal-to-noise ratio would collapse fast. You'd end up with a library where an early-draft scenario sits next to your official EOP, and Preppr couldn't tell the difference.</p><p>Keeping session attachments transient keeps your knowledge base clean. When something is ready to be part of your organization's permanent memory, you upload it deliberately. Until then, it lives where it belongs — in the session where you're working on it.</p><h2 id="how-it-powers-ask-preppr">How it powers Ask Preppr</h2><p>When you ask Ask Preppr a question, it doesn't just look at the document in front of you. It searches across your entire knowledge base — every doctrine document, every plan, every after-action report, every reference you've ever uploaded — and pulls the specific passages that matter for your question.</p><p>You see those passages cited directly in the answer. Every citation is clickable and opens to the exact section of the source document. Nothing is hallucinated; nothing is hidden. When Ask Preppr tells you what a plan requires or what a framework specifies, it's because Preppr Knowledge found it and showed you where it came from.</p><p>This is what makes Ask Preppr feel different from a generic chatbot. It isn't reasoning from a training-data approximation of what an emergency plan might say. It's reasoning from <em>your</em> knowledge base.</p><h2 id="how-it-powers-exercise-designer-v2">How it powers Exercise Designer V2</h2><p>Exercise Designer V2 leans on Preppr Knowledge in a different way. When you design an exercise, Preppr needs to ground the scenario in real doctrine, real capabilities, and real objectives. The knowledge base is where it gets that grounding.</p><p>If you ask for a tabletop exercise anchored state all-hazards plan and THIRA, Preppr Knowledge surfaces the relevant sections. If you're working an HSEEP-aligned exercise against specific core capabilities in our multi-year training plan, the knowledge base makes those capabilities available to the design process. Your previous exercises, your AARs, your improvement plans — all of it is available as raw material for the next exercise you build.</p><p>The practical effect is that exercises stop being generic templates with your logo on top. They become documents that reflect what your jurisdiction actually faces, what your team has actually learned, and what your doctrine actually requires.</p><h2 id="why-this-matters">Why this matters</h2><p>There's a pattern we see across the emergency management and public health field: organizations accumulate enormous amounts of knowledge — plans, AARs, SOPs, training materials, lessons from real incidents — and most of it sits unread on a shared drive. It's technically available, but practically inert. When a decision needs to be made or an exercise needs to be designed, nobody has time to dig through years of documents to find the piece that matters.</p><p>Preppr Knowledge is the answer to that problem. Your library stops being a graveyard and starts being a working memory. Everything you've learned is active, searchable, and automatically brought to bear on whatever you're doing right now.</p><p>This is the real shift from "AI chatbot" to "AI-native platform." A chatbot reads what you paste into it. A platform reads your organization.</p><h2 id="where-it-s-heading">Where it's heading</h2><p>At launch, Preppr Knowledge holds documents, profile data, and the national doctrine corpus — connected through the relationships Preppr can infer from that material. The roadmap expands what counts as a knowledge object, and the direction matters.</p><p><strong>Session conversations become knowledge objects.</strong> Right now, the work you do in Ask Preppr — the analysis you refine, the positions you work out, the decisions you arrive at — lives inside individual chat sessions. That's about to change. Your sessions will become first-class objects in Preppr Knowledge, connected to the documents they referenced, the plans they informed, and the decisions they supported. The thinking you did last month will be available to inform the work you're doing today. Preppr won't just remember your documents; it'll remember your reasoning, and know how that reasoning connects to everything else.</p><p><strong>Collaborators and reviewers become first-class knowledge.</strong> Preparedness is a team sport. The people who contributed to a plan, who reviewed an AAR, who signed off on an improvement action — these relationships matter. Preppr Knowledge will capture them as objects in the graph, so when you ask "who has weighed in on our mass casualty plan" or "which of our reviewers has flagged concerns about this scenario," you'll get an answer grounded in your actual working relationships.</p><p><strong>The graph deepens.</strong> As more object types come online — exercises, AARs, improvement actions, capabilities, threats, jurisdictions, resources, people — the relationships between them get richer. Preppr will be able to trace a lesson learned in 2023 forward through every plan revision, exercise, and policy change it influenced. It'll tell you which capabilities your last three exercises have and haven't touched. It'll answer questions like "show me every part of our preparedness posture that depends on a resource we flagged as at-risk six months ago" — questions that are impossible to answer against a pile of documents but straightforward against a connected model of your organization.</p><h2 id="the-substrate-not-the-surface">The substrate, not the surface</h2><p>There's a tendency in this industry to treat AI as a surface feature — a chat window bolted onto existing software. Preppr was built on the opposite premise: the AI is only as good as the knowledge it operates on, and knowledge is only useful when it's structured, connected, and alive.</p><p>Preppr Knowledge is that substrate. It's what makes Ask Preppr's answers trustworthy. It's what makes Exercise Designer V2's exercises real. And as it grows — from documents, to conversations, to collaborators, to a full relational graph of your preparedness posture — it's what will let Preppr do work that no document pile, no chatbot, and no bolted-on AI feature can touch.</p><p>The work of preparedness is the work of building, maintaining, and drawing on institutional knowledge. Preppr Knowledge is built to do exactly that.</p><h2 id="available-now">Available now</h2><p>Preppr Knowledge is live today in every Preppr Basic and Pro account at no additional cost. If you're already a subscriber, it's already working — every document you've uploaded and every profile field you've filled out is flowing into it automatically. The expansions described above — session-level memory, collaborator and reviewer knowledge, and the full relational knowledge graph — will roll out progressively over the coming releases.</p><p>Not a subscriber yet? <a href="https://app.preppr.ai/signup">Grab a 14-day, all access, free trial.</a></p>]]></content:encoded>
</item>
<item>
<title>How Preppr Designs Emergency Exercises That Actually Work</title>
<link>https://blog.preppr.ai/a/how-preppr-designs-emergency-exercises-that-actually-work/</link>
<guid isPermaLink="true">https://blog.preppr.ai/a/how-preppr-designs-emergency-exercises-that-actually-work/</guid>
<pubDate>Mon, 06 Apr 2026 12:00:00 GMT</pubDate>
<dc:creator>Justin Snair</dc:creator>
<category>Product Update</category>
<description>The original version of this article walked through eight sequential steps to build a tabletop exercise in Preppr. That was an accurate description of how V1 worked. Exercise…</description>
<content:encoded><![CDATA[<p><em>Updated for Exercise Designer V2</em></p><p>The original version of this article walked through eight sequential steps to build a tabletop exercise in Preppr. That was an <a href="https://blog.preppr.ai/a/how-preppr-ai-creates-emergency-exercises-that-actually-work/">accurate description of how V1</a> worked. Exercise Designer V2 is a different product with a different philosophy, and it deserves a different explanation.</p><p>The core change: <a href="https://preppr.ai/feature-exdesigner">Exercise Designer V2</a> is not a linear wizard. It meets you wherever you are.</p><p>Come with nothing — Preppr builds from scratch. Come with a rough scenario concept — Preppr picks it up and runs. Come with a fully drafted SITMAN covered in reviewer comments and redlines — Preppr adjudicates the feedback, preserves what should be preserved, maps downstream changes, and gets you to a finished exercise plan. The starting point doesn't determine the quality of the output. The ending point is always the same: a complete, doctrine-aligned, delivery-ready exercise.</p><p>Here's what V2 can do and why it's designed that way.</p><h2 id="four-entry-modes-one-destination">Four entry modes, one destination</h2><p>Every design session begins with Discovery. Preppr consults your source documents — Emergency Operations Plans, threat assessments, AARs, THIRAs, mutual aid agreements — surfaces what it found, and presents a starting point for your review. From there, you choose how much you want to drive and how much you want to hand off. Four modes cover the full range.</p><p><strong>Describe</strong> is for planners who want to stay in control. You provide the direction — scenario concept, audience, threat, objectives, constraints, any combination — and Preppr builds around what you've given it. Describe what you know, skip what you don't, and Preppr fills the gaps without overwriting your intent.</p><p><strong>Recommend</strong> surfaces Preppr's own read of your materials as a starting point. Based on what your documents reveal — gaps from prior AARs, risks flagged in your THIRA, capabilities that haven't been tested — Preppr proposes a scenario, objectives, and structure. You review and confirm before anything is built.</p><p><strong>Autopilot</strong> activates after Discovery. Every design session begins with a Discovery phase — Preppr analyzes your source documents, infers the right hazard, audience, capabilities, and objectives, and surfaces a Discovery Summary for your review. From there, you can take over or hand off. Click Autopilot and Preppr runs the complete generation pipeline end-to-end without you: scenario narrative, objectives, modules, injects, discussion questions, controller notes, and MSEL.</p><p>What makes Autopilot distinctive is how it makes decisions. Rather than silently generating content, Preppr deliberates visibly with a synthetic persona that you define — a simulated practitioner who represents your organizational context, challenges design choices, demands justification grounded in your source documents, and pushes back when something doesn't fit. You control how demanding that persona is. The result is a design that's been stress-tested against your own organizational reality before you ever open it. Come back to a review-ready exercise plan. Revise what needs adjustment, approve what doesn't.</p><p><strong>Revise</strong> is for organizations that already have exercise materials. It accepts SITMANs filled with redlines and reviewer comments, adjudicates the feedback, preserves existing content, and maps downstream dependencies before presenting a proposed change plan. After human approval, Preppr updates the exercise and produces new documentation along with a crosswalk that records every change accepted, rejected, and the reasoning behind each decision. What most organizations manage through email chains and tracked-changes documents becomes a structured, auditable process.</p><h2 id="review-is-part-of-the-design">Review is part of the design</h2><p>Exercise design isn't just content production. It's a negotiation — between planners, reviewers, subject matter experts, and leadership — and the work of managing that negotiation typically falls outside any tool. Emails accumulate. Tracked-changes documents multiply. Feedback arrives in different formats from people with different levels of authority, and someone has to reconcile it all, document what was taken, explain what wasn't, and get sign-off before the exercise goes to participants.</p><p>Preppr builds review into the design process at every stage, not just at the end.</p><p>Drafts can be shared with reviewers at any point — after scenario confirmation, after injects are generated, after the full plan is assembled. Reviewers receive a structured interface to comment, redline, and submit feedback directly. When submissions come in, Preppr consolidates them, identifies conflicts, and generates a proposed resolution plan. You review and approve. Preppr updates the content and produces a crosswalk that records every change: what was accepted, what was rejected, and the reasoning behind each decision.</p><p>That crosswalk isn't administrative overhead — it's the documentation that makes the exercise defensible. When leadership asks why a particular scenario element was included or dropped, the answer is already written. When an accreditor wants evidence of a structured planning process, it exists. When you're designing next year's exercise and want to know what this year's reviewers pushed back on, it's in the record.</p><p>The result is that stakeholder buy-in and exercise quality reinforce each other rather than trading off. Getting more reviewers involved doesn't mean more reconciliation work. It means a more thoroughly vetted exercise and a more complete audit trail.</p><h2 id="what-drives-the-design-three-intelligence-layers">What drives the design: three intelligence layers</h2><p>Regardless of which entry mode you use, every exercise Preppr designs draws from the same three sources of intelligence simultaneously.</p><p><strong>Your documents.</strong> Preppr reads your uploaded organizational materials and uses them throughout design — not just at the start. When generating injects, it references your actual response timelines, resource inventories, and documented gaps. When writing controller notes, it cites your SOPs and plan sections. When identifying objectives, it surfaces improvement areas from your own AARs. The exercise reflects your organizational reality, not a generic template. Your entire document library is available within the Exercise Designer v2.</p><p>You can also work with your documents in Ask Preppr. There you can conduct analyzes, develop content, and easily pull it into the Exercise Designer.</p><p><strong>Your organizational substrate.</strong> Documents are a starting point. Preppr Knowledge is what they become.</p><p>Every time you work with Preppr — uploading plans, running exercises, completing assessments, engaging stakeholders through Collaborate — that information doesn't sit inert in a file store. It's processed into a relational knowledge graph: a structured, living representation of your organization's preparedness architecture. Your facilities and jurisdictions. Your capabilities and the gaps in them. Your inter-agency relationships and the agreements that govern them. The findings from your exercises and what you did — or didn't do — about them. Preppr Knowledge builds this picture continuously, from every source Preppr touches.</p><p>This matters because most preparedness information exists in isolation. A plan references a resource that appears nowhere else. An AAR finding gets filed and forgotten. A THIRA gap never makes it into an exercise objective. Preppr Knowledge closes those loops. When Preppr designs an exercise, it's not just reading your documents — it's querying a connected model of your organization that knows how your capabilities relate to your risks, how your plans relate to each other, and where the gaps in that picture are. The exercise that comes out reflects not just what your documents say, but what the pattern of your documents reveals.</p><p>Preppr Knowledge also compounds. The first exercise you build with Preppr is informed by what you've uploaded. The fifth is informed by four cycles of exercise history, facilitated discussions, improvement plan tracking, and continuous intelligence feeds from Preppr Intelligence and Atlas. The substrate grows with your program. What you learn doesn't disappear between exercises — it becomes the foundation the next one is built on.</p><p><strong>Open source intelligence.</strong> When you start an exercise design, Preppr Intelligence automatically pulls real-time threat data from open sources — current incident reporting, active advisories, recent events matching your hazard type and geography. This flows directly into scenario generation, grounding your narrative in what's actually happening rather than what happened in a training scenario five years ago. The result is a scenario that feels current, locally relevant, and professionally informed without requiring the planner to separately research and synthesize threat data.</p><p><strong>Doctrine and framework knowledge.</strong> V2 is built on a knowledge graph that encodes FEMA's Community Lifelines, FEMA Core Capabilities, HSEEP evaluation methodology, and crosswalks to PHEP, HPP, EPHEPR, and Joint Commission EM standards. This isn't a lookup table — it's a connected structure that drives design decisions throughout. When you select a hazard, the system understands which lifelines are at risk, which capabilities should be tested, and how those capabilities map across frameworks. Objectives connect to capabilities. Injects connect to objectives. Controller notes connect to evaluation criteria. Every element of the exercise traces back to doctrine, automatically.</p><h2 id="scenario-design-grounded-not-generic">Scenario design: grounded, not generic</h2><p>V2 includes a library of 76+ scenario starter cards, each grounded in real incidents and tagged to specific Community Lifelines and capability frameworks. These aren't generic "earthquake in a major city" prompts — they encode the operational pressure points, infrastructure dependencies, and coordination challenges that have actually emerged in similar events.</p><p>Each starter card includes a severity emergence framework — a structured model of how conditions escalate across incident phases. Rather than designing a scenario at a fixed severity level, you're designing a scenario that evolves realistically: what's true at hour two is different from what's true at hour twelve, and the inject sequence reflects that progression. Participants face compounding decisions, not isolated ones.</p><p>For multi-agency and multi-jurisdictional exercises, V2 supports Tabletop with Discussion Groups — a full breakout-to-plenary structure where functional groups discuss group-specific questions tied to each inject's stress mapping, then reconvene for facilitated synthesis. Each inject is mapped to which groups it primarily stresses, which it affects secondarily, and which it leaves peripheral. Questions are weighted accordingly. Plenary prompts target the cross-group tensions that matter most. The design produces an exercise that actually tests coordination, not just individual group knowledge.</p><h2 id="the-design-to-delivery-pipeline">The design-to-delivery pipeline</h2><p>Exercise Designer V2 doesn't produce a document — it produces a deliverable. The exercise plan that comes out of the designer feeds directly into Preppr Exercise for live facilitation. Discussion questions become the AI facilitator's prompts. Module structure becomes the delivery sequence. Controller notes become real-time guidance. The crosswalk from Revise mode becomes part of the exercise record.</p><p>This matters because most of the administrative cost in exercise programs comes from translation work — taking a design document and converting it into a facilitator guide, a participant handout, a controller runsheet. In V2, that translation doesn't happen because the design and the delivery artifact are the same thing.</p><p>For organizations that want to ground their exercise design in genuine stakeholder intelligence before any of this begins, Preppr Collaborate provides the upstream layer. Through a structured four-phase workflow — Campaign Setup, Contributions, Synthesis, and Integration — Collaborate aggregates input from anywhere between a handful of stakeholders and a community-wide engagement of thousands. The synthesis engine identifies consensus patterns, critical dependencies, and gaps, then generates draft inject recommendations that feed directly into Exercise Designer. The result is an exercise that reflects what your stakeholders actually know and need, not what a single planner assumed.</p><h2 id="why-non-linear-matters">Why non-linear matters</h2><p>The old model of exercise design assumed a clean starting point: blank page, clear scope, dedicated planning time. Most organizations don't have that. They have a draft from last year with comments that never got addressed. They have a new threat that needs to be incorporated into an existing framework. They have a planning committee that can't agree on scope. They have a grant requirement with a 30-day deadline.</p><p>V2 is designed for those conditions. The starting point is wherever you actually are. The output is always a complete, doctrine-aligned exercise plan — with full documentation of how it was built and what changed along the way.</p><p>That audit trail isn't a bureaucratic artifact. It's what makes the exercise defensible to leadership, useful for accreditation, and valuable as a baseline for the next cycle.</p><p><em>Ready to see what Exercise Designer V2 can do with your materials?</em> <a href="https://app.preppr.ai/signup"><em>Start a free trial →</em></a> <em>or</em> <a href="https://preppr.ai/demo"><em>book a demo →</em></a></p>]]></content:encoded>
</item>
<item>
<title>How to Run Tabletop Exercises That Set You Up for AAR Success</title>
<link>https://blog.preppr.ai/a/how-to-run-tabletop-exercises-that-set-you-up-for-aar-success/</link>
<guid isPermaLink="true">https://blog.preppr.ai/a/how-to-run-tabletop-exercises-that-set-you-up-for-aar-success/</guid>
<pubDate>Sun, 05 Apr 2026 12:00:00 GMT</pubDate>
<dc:creator>Justin Snair</dc:creator>
<category>General</category>
<description>A tabletop exercise can feel like the ’easy’ kind of exercise: no deployments, no field gear, no logistics gauntlet. Yet many organizations still dread the part that comes…</description>
<content:encoded><![CDATA[<p>A tabletop exercise can feel like the "easy" kind of exercise: no deployments, no field gear, no logistics gauntlet. Yet many organizations still dread the part that comes after — the after-action report (AAR). Not because the writing is inherently hard, but because the exercise didn't generate clean, defensible evidence in the first place.</p><p>If you want to AAR success, the highest-leverage move is to run tabletop exercises that are designed for evaluation and documentation from the start. The good news: you do not need a bigger team or more hours. You need tighter intent, better capture, and a workflow that turns decisions into findings.</p><p>This guide covers both the manual discipline behind high-quality exercise documentation and how purpose-built platforms like Preppr can compress the timeline across every phase — from design to delivery to the final AAR/IP.</p><h2 id="why-aars-drag-and-why-they-disappoint">Why AARs drag (and why they disappoint)</h2><p>Most slow AAR cycles trace back to a few predictable failure modes:</p><ul><li>Objectives are too broad ("test communications") or too many, so notes sprawl and analysis becomes subjective.</li><li>No evaluation plan exists beyond "take notes," so the team cannot distinguish a meaningful gap from a normal discussion tangent.</li><li>Decisions are not time-stamped or attributed, which forces the author to reconstruct the story later from memory.</li><li>Key artifacts are missing (situation updates, draft messages, resource requests, policy calls), leaving the AAR heavy on opinions and light on evidence.</li><li>Corrective actions are vague ("improve coordination"), which makes the AAR easy to finalize but useless to implement.</li></ul><p>FEMA's Homeland Security Exercise and Evaluation Program (HSEEP) has emphasized this for years: evaluation and improvement planning are not an afterthought — they are part of the exercise lifecycle. If your organization uses HSEEP (or borrows its logic), your tabletop exercise should be engineered to produce an AAR/IP with minimal friction.</p><p>The root problem is that most organizations treat exercise design, exercise delivery, and exercise documentation as three separate projects. They are not. They are one continuous workflow, and the quality of the final AAR is determined almost entirely by decisions made in the first two phases.</p><h2 id="phase-1-design-for-documentation-from-the-start">Phase 1: Design for documentation from the start</h2><h3>Write fewer, sharper objectives — and define what "good" looks like</h3><p>Strong objectives are specific enough to evaluate and narrow enough to capture. A practical target for most tabletop exercises is 3 to 5 objectives.</p><p>Instead of writing objectives that describe topics, write objectives that describe observable performance.</p><p><strong>Example shift:</strong></p><ul><li>Topic-style: "Discuss information sharing."</li><li>Performance-style: "Demonstrate how the organization validates, approves, and disseminates public information within X minutes across identified channels, including accessibility and language considerations."</li></ul><p>You do not need to force numeric targets into every objective, but you do need observable criteria. Otherwise, your AAR becomes a debate.</p><h3>Decide what evidence you will accept</h3><p>Before the exercise, define what counts as evidence for each objective. Evidence can be:</p><ul><li>A decision ("We will activate the JIC at 0900, led by…")</li><li>An artifact (draft warning message, ICS 214 entry, resource request)</li><li>A demonstrated process (who approves what, what tool is used, where it is documented)</li></ul><p>A simple mapping prevents "blank page AAR syndrome":</p><table><thead><tr><th>Objective</th><th>Evidence to capture</th><th>Primary capture method</th><th>Owner</th></tr></thead><tbody><tr><td>Public information approval and release</td><td>Time of decision, approver, channels, draft message text, accessibility checks</td><td>Time-stamped decision log + artifact collection</td><td>Scribe + PIO evaluator</td></tr><tr><td>Resource prioritization</td><td>Decision criteria, competing demands, unmet needs, mutual aid triggers</td><td>Decision log + evaluator notes tied to objective</td><td>Operations evaluator</td></tr><tr><td>Cyber incident coordination</td><td>Escalation path, containment decisions, legal/privacy constraints, vendor contact</td><td>Decision log + artifact capture</td><td>IT/security evaluator</td></tr></tbody></table><h3>Build an evaluation structure that matches your reality</h3><p>HSEEP provides a formal structure (capabilities, critical tasks, performance measures). Some organizations adopt it fully. Others adapt it.</p><p>Either way, pick a structure your team will actually use — capability-based, function-based, or risk-based. The best structure is the one that lets evaluators tag observations consistently, so analysis writes itself later.</p><h3>Build your injects around decision points, not discussion topics</h3><p>The most common tabletop design failure is structuring modules around topics instead of forcing decisions. "Discuss your communications approach" produces meandering conversation. "You have 47 confirmed measles cases in three schools, regional hospital is at 95% capacity, and your public health director is asking whether to activate the JIC — what do you do?" produces a decision.</p><p>Design each inject to force a specific, evaluable choice:</p><ul><li>"Do you activate? At what level? Who has authority?"</li><li>"What is your public message, and who approves it?"</li><li>"What mutual aid triggers apply, and what information is required?"</li><li>"What do you stop doing to free capacity?"</li></ul><p>When your scenario is structured this way, your AAR analysis becomes straightforward because you have comparable decision moments across exercises.</p><h3>How Preppr accelerates exercise design</h3><p>Exercise design is where most organizations spend the most time and where the most institutional knowledge gets lost. <a href="https://preppr.ai/feature-exdesigner">Preppr's Exercise Designer</a> addresses this directly.</p><p>The platform offers four design entry modes — Describe, Recommend, Autopilot, and Revise — so you can generate a complete exercise framework in minutes rather than days, starting from whatever level of detail you have. A library of 76+ scenario starter cards grounded in real incidents gives planners a defensible, doctrine-aligned foundation rather than a blank page.</p><p>Revise mode deserves particular attention for organizations that already have existing exercise materials. It accepts SITMANs filled with redlines and reviewer comments, adjudicates the feedback, preserves existing content, and maps downstream dependencies before presenting a proposed change plan. After human approval, Preppr updates the exercise and produces new documentation along with a crosswalk that records every change accepted, rejected, and the reasoning behind each decision — creating an auditable revision history that most organizations currently manage through email threads and tracked-changes documents.</p><p>When you start an exercise design, Preppr Intelligence automatically pulls real-time threat intelligence from open sources — current incident data, active advisories, recent events matching your hazard type and geography. This context flows directly into scenario generation, grounding your narrative in what's actually happening rather than generic templates. The result is a scenario that feels current, locally relevant, and professionally informed without requiring the planner to manually research and synthesize threat data before they can start designing.</p><p><strong>Autopilot</strong> is Preppr's fully automated exercise design mode. Point it at your source documents — threat assessments, after-action reports, THIRAs, planning documents — and walk away. Autopilot analyzes your materials, infers the right hazard, audience, capabilities, and objectives, and presents a Discovery Summary for your review before it builds. Once you confirm, it runs the complete generation pipeline end-to-end: scenario narrative, objectives, modules, injects, discussion questions, controller notes, and MSEL. Come back to a complete, review-ready exercise plan. From there, you revise what needs refinement and approve what doesn't.</p><p>The external review workflow extends this across stakeholders. In-progress or final drafts can be sent to reviewers who can edit, comment, and submit feedback directly.</p><p>Preppr receives the submissions, analyzes them, generates a consolidation plan, and — with your approval — updates the content and produces the crosswalk. What typically takes weeks of reconciliation across multiple reviewers becomes a structured, documented process.</p><p>More broadly, Preppr structures every exercise around HSEEP-aligned objectives, functional groups, and capability frameworks from the start. The inject sequence is built to force evaluable decisions — not just discussion topics. Because the design layer is built for documentation, the evaluation structure is already in place before the first participant logs in.</p><p>Preppr also maps exercises to FEMA Community Lifelines and FEMA Core Capabilities, ensuring that every tabletop generates findings that connect to the planning frameworks your leadership and funders expect to see.</p><p>For organizations that want to ground their exercise design in genuine stakeholder intelligence rather than a single planner's assumptions, <a href="https://preppr.ai/feature-collaborate">Preppr Collaborate</a> provides the foundation. Through a structured four-phase workflow — Campaign Setup, Contributions, Synthesis, and Integration — Collaborate aggregates input from anywhere between a handful of stakeholders and a community-wide engagement of thousands. Organizations contribute their capabilities, vulnerabilities, and operational context through guided conversations. The synthesis engine identifies consensus patterns, critical dependencies, and gaps across all submissions, then generates draft inject recommendations that feed directly into exercise design. The result is an exercise grounded in what your stakeholders actually know about their own organizations — not a scenario built in isolation.</p><h2 id="phase-2-deliver-the-exercise-to-generate-usable-findings">Phase 2: Deliver the exercise to generate usable findings</h2><h3>Staff for documentation, not just facilitation</h3><p>AAR speed improves dramatically when roles are clear and resourced. Facilitator, evaluator, and scribe are different jobs — and many tabletop exercises overload one person with all three.</p><p>At minimum:</p><ul><li><strong>Facilitator</strong> keeps pace, enforces scope, and drives decision points.</li><li><strong>Scribe</strong> runs a time-stamped decision log and captures key artifacts.</li><li><strong>Evaluator(s)</strong> focus on specific objectives and record observations in a structured way.</li></ul><p>If you are short-staffed, protect the scribe role. A strong scribe is often the difference between a two-day AAR and a two-week AAR.</p><h3>Use a decision log as the backbone</h3><p>AARs are easiest to write when you can reconstruct the exercise as a sequence of decision points. A practical decision log captures:</p><ul><li>Time (real or simulated)</li><li>Decision required</li><li>Options considered</li><li>Decision made and owner</li><li>Constraints and assumptions</li><li>Follow-up actions created</li></ul><p>This becomes your narrative — and it becomes your improvement plan seed.</p><h3>Set ground rules that support clean evaluation</h3><p>A few ground rules improve both psychological safety and documentation quality:</p><ul><li><strong>No-fault learning environment</strong>: focus on systems and decisions, not personal performance.</li><li><strong>Assumptions are allowed, but must be stated</strong>: the scribe captures them, because assumptions often become AAR findings.</li><li><strong>When a decision is made, it is logged</strong>: the facilitator pauses briefly to confirm wording.</li></ul><p>This does not slow the room if you do it consistently.</p><h3>Use structured capture buckets</h3><p>A simple rule: every observation should land in one of a few categories.</p><table><thead><tr><th>Capture bucket</th><th>What it is</th><th>Why it speeds the AAR</th></tr></thead><tbody><tr><td>Decision</td><td>A choice made, by someone, at a time</td><td>Becomes the narrative backbone</td></tr><tr><td>Observation</td><td>What was done well or poorly, tied to an objective</td><td>Turns into strengths and areas for improvement</td></tr><tr><td>Gap</td><td>A missing process, unclear authority, broken dependency</td><td>Becomes a finding with a clear "what failed"</td></tr><tr><td>Root cause hint</td><td>Why the gap exists (training, policy, tools, staffing)</td><td>Prevents shallow corrective actions</td></tr><tr><td>Action item</td><td>A proposed fix with an owner concept</td><td>Seeds the improvement plan immediately</td></tr></tbody></table><p>If evaluators tag notes this way while the exercise runs, your AAR writing becomes compilation and editing, not interpretation.</p><h3>Capture artifacts in the moment (even if they are imperfect)</h3><p>Artifacts are the fastest path to a defensible AAR. In a tabletop exercise, artifacts can be drafts and placeholders:</p><ul><li>A draft public message or internal notification</li><li>A screenshot of the situation status format participants agree to use</li><li>A filled-in resource request template</li><li>A call-down list with corrections marked</li></ul><p>If your organization has templates, use them during the exercise. If you do not, this is your signal to create lightweight templates as part of improvement planning.</p><h3>Separate parking lot issues from evaluable findings</h3><p>Tabletop exercises often surface important but non-evaluable issues (budget constraints, long-term staffing shortages, vendor dissatisfaction). Capture them, but do not let them clog the AAR. Use a parking lot log with two fields: "topic" and "next step owner."</p><h3>How Preppr facilitates and captures in real time</h3><p>This is where Preppr's end-to-end delivery capability changes the calculus for under-resourced teams.</p><p><a href="https://preppr.ai/feature-exercise">Preppr Exercise</a> guides participants through every phase of exercise delivery — Welcome and check-in, Exercise Overview, Scenario Presentation, Module Discussion, Group Report-Out, Cross-Group Synthesis, Final Report, Hotwash, and Improvement Planning — with purpose-built prompting for each phase. The facilitator doesn't need to hold all of this in their head or manage a complex runsheet; Preppr handles the structure.</p><p>During module discussions, Preppr captures a <strong>response synthesis</strong> for every question — a 2-4 sentence approved summary of what the group decided, validated by participants before moving forward. This is the documentation artifact that makes post-exercise report writing trivial: the "final answer" to every question is already written, reviewed, and attributed.</p><p>Preppr also tracks document intelligence during discussions. When uploaded plans are available, the system identifies gaps between what participants say they would do and what their documented procedures actually specify — providing the kind of evidence-based findings ("Group stated they would use phone calls to request mutual aid, unaware that the Regional Mutual Aid Agreement Section 2.3 requires WebEOC-based requests with 4-hour response commitments") that practitioners normally reconstruct hours after the exercise ends.</p><p>Multi-group exercises benefit from Preppr's cross-group synthesis capability, which surfaces coordination gaps and conflicting assumptions across functional groups — the systemic findings that single-group documentation almost always misses.</p><p>Throughout delivery, the decision log, observations, gaps, root cause hints, and action items are being built automatically in the background. The scribe role still matters, but the platform is doing the heavy lifting on structure.</p><h2 id="phase-3-document-debrief-and-improve">Phase 3: Document, debrief, and improve</h2><h3>Make the hotwash a draft AAR, not a group therapy session</h3><p>A hotwash can either accelerate the AAR or derail it. The difference is facilitation discipline.</p><p>Instead of asking "how did it go," use objective-anchored prompts:</p><ul><li>"For Objective 1, what specifically worked as designed?"</li><li>"Where did you lose time, and why?"</li><li>"What decision authority was unclear or contested?"</li><li>"What information did you wish you had, and who owns it?"</li></ul><p>Convert discussion into draft corrective actions on the spot. You are not finalizing wording in the room — you are capturing:</p><ul><li>Corrective action (what will change)</li><li>Owner (person or role)</li><li>Due date concept (near, mid, long)</li><li>Validation method (how you will test the fix)</li></ul><p>This turns the AAR from a document into a management tool.</p><h3>Write a faster, better AAR using an evidence-first structure</h3><p>If you followed the practices above, AAR writing becomes mostly assembly.</p><p><strong>A practical structure that aligns with HSEEP-style expectations:</strong></p><p><em>Executive summary that is honest and specific.</em> Decision-makers read this. Avoid generic statements. Include the top strengths worth sustaining, the highest-risk gaps, and any cross-cutting systemic issues (approval latency, vendor dependencies, unclear roles).</p><p><em>Analysis organized by objective, with evidence.</em> For each objective: what was expected, what happened (evidence-based), strengths, areas for improvement, root cause themes.</p><p>The key is to write findings as evidence statements:</p><blockquote>"The organization did not have a documented threshold for escalating to the JIC, resulting in a 25-minute debate and inconsistent messaging drafts across units."</blockquote><p>That is a finding you can act on.</p><p><em>Improvement plan that is implementable.</em> AARs fail when corrective actions are unowned or untestable. At minimum, every action needs: corrective action, responsible party, timeline, and how completion will be validated.</p><h3>Capability ratings that drive accountable improvement planning</h3><p>HSEEP's AAR/IP approach uses a four-level performance rating system for each capability tested:</p><ul><li><strong>P</strong> (Performed without Challenges): Objectives achieved without notable issues</li><li><strong>S</strong> (Performed with Some Challenges): Objectives achieved with notable challenges</li><li><strong>M</strong> (Performed with Major Challenges): Objectives barely achieved due to significant issues</li><li><strong>U</strong> (Unable to be Performed): Objectives could not be achieved</li></ul><p>These ratings only hold up when they are tied to specific evidence from the exercise. A capability rated "S" should come with a clear explanation of what the challenge was — otherwise leadership has no basis for prioritizing the improvement investment.</p><h3>Avoid the most common slow-AAR traps</h3><table><thead><tr><th>Trap</th><th>What it looks like</th><th>Better practice</th></tr></thead><tbody><tr><td>The AAR tries to cover everything</td><td>Every conversation becomes a finding</td><td>Limit scope to objectives, park the rest</td></tr><tr><td>Notes are unstructured</td><td>Pages of narrative with no tags</td><td>Use decision log + objective-tagged observations</td></tr><tr><td>Findings lack root cause</td><td>"Need more training" repeated everywhere</td><td>Capture constraints and why decisions were hard</td></tr><tr><td>Corrective actions are not testable</td><td>"Improve coordination"</td><td>Define a change plus how it will be validated</td></tr><tr><td>No governance after publication</td><td>AAR is filed and forgotten</td><td>Assign owners and review progress on a cadence</td></tr></tbody></table><h3>How Preppr generates the AAR/IP</h3><p>By the time Preppr's exercise delivery workflow reaches the Improvement Planning phase, the system already has:</p><ul><li>Approved response syntheses for every question in every module</li><li>Module report-outs with POETE analysis (Plans, Organization, Equipment, Training, Exercise needs)</li><li>Cross-module synthesis identifying systemic patterns</li><li>P/S/M/U capability ratings with evidence</li><li>Hotwash findings organized by strength and gap</li><li>Draft improvement actions with SMART framing</li></ul><p>The Improvement Planning session uses this pre-generated material as a structured starting point. Draft actions are presented to the group for collaborative refinement — owners, timelines, success criteria, blockers, and validation methods are confirmed in the room. The resulting improvement plan isn't a wish list; it's a set of explicit commitments with names attached.</p><p>The documentation that comes out of this process reflects everything captured across the full exercise lifecycle. Because every section traces back to validated participant responses, facilitators aren't writing an interpretation of what happened — they're assembling documentation that participants already approved.</p><h2 id="turn-aars-into-a-continuous-learning-cycle">Turn AARs into a continuous learning cycle</h2><p>AAR speed matters, but what matters more is whether the organization actually improves.</p><p>A lightweight governance loop that works across emergency management, healthcare, and private-sector resilience teams:</p><ol><li><strong>Publish the AAR quickly</strong> while memory is fresh (even if some sections are "draft pending validation").</li><li><strong>Hold a corrective action review</strong> with leadership to approve owners and timelines.</li><li><strong>Track corrective actions like operational work</strong>, not like training paperwork.</li><li><strong>Re-test the highest-risk fixes</strong> in the next tabletop exercise, so improvements are validated, not assumed.</li></ol><p>When you do this, tabletop exercises stop being annual events and start being part of operational readiness.</p><p>The compounding effect is significant. Each exercise cycle produces an approved improvement plan. The next exercise tests whether those improvements worked. Preppr's knowledge graph — Preppr Bedrock — captures organizational intelligence from every exercise cycle, so institutional memory doesn't walk out the door when personnel turn over and every subsequent exercise starts from a richer baseline.</p><h2 id="the-simplest-takeaway">The simplest takeaway</h2><p>If you want faster, better AARs, stop treating the AAR like a writing task. Treat it like the output of a well-instrumented exercise lifecycle.</p><p>Clear objectives, disciplined decision capture, artifact collection, and objective-tied hotwash prompts will do more for your AAR quality than any post-exercise writing sprint. And once your AARs consistently produce implementable improvement plans, your tabletop exercise program becomes a true readiness engine — not a compliance checkbox.</p><p>For organizations looking to compress the timeline across all three phases — design, delivery, and documentation — <a href="https://preppr.ai/">Preppr</a> is built around this end-to-end lifecycle. The platform handles the structure so practitioners can focus on the judgment that only they can provide: the decisions, the tradeoffs, and the organizational context that no AI can substitute for.</p><p><a href="https://blog.preppr.ai/a/how-preppr-ai-creates-emergency-exercises-that-actually-work/"><strong>Explore Preppr's exercise design workflow →</strong></a>  <a href="https://preppr.ai/feature-exercise"><strong>See how Preppr delivers and documents tabletop exercises →</strong></a></p>]]></content:encoded>
</item>
<item>
<title>10 Questions (and 7 Buzzwords) to Ask Any AI Vendor Before You Sign</title>
<link>https://blog.preppr.ai/a/10-questions-and-7-buzzwords-to-ask-any-ai-vendor-before-you-sign/</link>
<guid isPermaLink="true">https://blog.preppr.ai/a/10-questions-and-7-buzzwords-to-ask-any-ai-vendor-before-you-sign/</guid>
<pubDate>Sat, 28 Mar 2026 12:00:00 GMT</pubDate>
<dc:creator>Justin Snair</dc:creator>
<category>General</category>
<description>AI is transforming emergency management and business continuity — and vendors are racing to claim the space. Some are building real capability. Others are wrapping ChatGPT in…</description>
<content:encoded><![CDATA[<p>AI is transforming emergency management and business continuity — and vendors are racing to claim the space. Some are building real capability. Others are wrapping ChatGPT in a branded interface and calling it innovation.</p><p>As someone who's spent twenty years in this field and now builds AI tools for it, I've watched (and experienced) procurement conversations up close. I've heard about and seen what vendors promise in demos and what their contracts actually say. I've talked to buyers who signed and regretted it, and buyers who asked the right questions and avoided costly mistakes.</p><p>Here are ten questions and seven buzzwords that will separate substance from marketing — regardless of which vendor you're evaluating, including us.</p><h2 id="1-does-your-contract-match-what-you-just-told-me">1. Does your contract match what you just told me?</h2><p>This is the most important question on this list, and it should be the first thing you check.</p><p>Ask the vendor in the demo: <em>Do you train your AI models on our data?</em> Write down exactly what they say. Then read the contract. Read the data rights section, the license grants, the intellectual property clause. Compare the words on the page to the words in the room.</p><p>If they don't match, walk away. A company that says one thing in a meeting and writes another in the contract is telling you who they are.</p><h2 id="2-what-happens-to-my-data-if-i-leave">2. What happens to my data if I leave?</h2><p>Every vendor will tell you they don't lock you in. Ask them to prove it. Can you export all of your data — documents, outputs, findings, exercise records, user activity — in a format you can actually use? How long do they retain your data after termination? Is deletion automatic, or do you have to request it? What's the timeline?</p><p>"You can cancel anytime" is not the same as "you can leave with everything you brought and everything you built."</p><h2 id="3-if-i-contribute-content-to-a-shared-library-can-i-fully-retract-it">3. If I contribute content to a shared library, can I fully retract it?</h2><p>Some platforms aggregate user-contributed content into shared knowledge bases. This can be valuable — but it raises a critical question about data recall.</p><p>If you upload a document to a shared library and later revoke permission, does the vendor delete only your original file, or do they also remove every derivative output that referenced your content in other users' sessions? If they can't recall derivative content, your data lives on in the system long after you've "deleted" it. Ask for specifics. "You can remove your documents at any time" is not the same thing as full retraction.</p><h2 id="4-can-i-see-the-methodology-behind-your-evaluations">4. Can I see the methodology behind your evaluations?</h2><p>If a vendor tells you their tool evaluates your plans and gives you a score or a rating, ask what it's evaluating against. Is it a checklist? A compliance framework? A proprietary rubric? Can you see it?</p><p>A red-yellow-green scorecard is only useful if you understand the standard behind it. If the methodology is opaque — if the tool tells you your plan is "compliant" without explaining what compliance means in this context — then the output is decoration, not analysis. You need to know what's being measured, against what standard, and whether that standard is appropriate for your organization.</p><h2 id="5-what-does-your-tool-actually-produce-and-what-do-i-still-have-to-do-myself">5. What does your tool actually produce — and what do I still have to do myself?</h2><p>This is where demos get slippery. A tool that "designs exercises" might mean it generates scenario text you have to copy, paste, format, and organize into a situation manual yourself. It's a lot like what ChatGPT or Claude might do.</p><p>Or it might mean it produces a complete, structured, HSEEP-aligned exercise plan with objectives, injects, discussion questions, facilitation guidance, and exportable documentation.</p><p>Those are very different products at very different price points. Ask to see the actual output. Not a slide about the output. The output itself. Then ask: what do I have to do between this output and a deliverable I can hand to my boss?</p><h2 id="6-does-your-tool-handle-the-full-workflow-or-just-one-piece-of-it">6. Does your tool handle the full workflow, or just one piece of it?</h2><p>Exercise design is not a single task. It's a workflow that includes document analysis, stakeholder engagement, threat intelligence gathering, scenario development, inject sequencing, facilitation planning, exercise delivery, observation capture, after-action reporting, and improvement planning.</p><p>Ask the vendor which of these steps happen inside their platform and which ones you're doing outside it — in Word, in email, in meetings, by hand. A tool that handles one step well but leaves you to manually manage the other nine is not a platform. It's a feature.</p><h2 id="7-how-does-your-tool-handle-stakeholder-review-and-revision">7. How does your tool handle stakeholder review and revision?</h2><p>Exercise content rarely survives first contact with stakeholders. Situation manuals get reviewed, commented on, revised, and re-reviewed — sometimes by dozens of people across multiple agencies.</p><p>Ask the vendor: does this process happen inside your tool, or does it happen in email and tracked-changes Word documents? If a stakeholder submits fifty comments on a draft, does the tool help you adjudicate them, or are you reading them one by one and making changes manually? The review and revision cycle is where most exercise timelines die. If the tool doesn't address it, you haven't saved as much time as the demo suggested.</p><h2 id="8-can-i-get-a-free-trial-with-my-own-data">8. Can I get a free trial with my own data?</h2><p>Not a guided demo with the vendor's sample data. A free trial where you upload your own plans, work with your own scenarios, and evaluate the tool against your actual workflow.</p><p>If the answer is no — if the vendor requires a contract before you can test the product with real-world content — ask yourself why. A product that can't survive contact with your reality is a product that's selling its demo, not its capability.</p><h2 id="9-who-built-this-and-do-they-understand-my-work">9. Who built this, and do they understand my work?</h2><p>AI is a powerful technology, but it's a means, not an end. The quality of an AI tool for emergency management depends entirely on whether the people who built it understand emergency management — the doctrine, the workflows, the politics, the constraints, the difference between compliance theater and genuine readiness.</p><p>Ask who's on the team. Ask about their background in the field. Ask them to explain HSEEP, or core capabilities, or the difference between a tabletop discussion and a full-scale exercise. If the answers are vague, the product probably is too.</p><h2 id="10-what-does-your-tool-look-like-after-the-first-use">10. What does your tool look like after the first use?</h2><p>The best demo is always the first one. The real question is what happens on day thirty, day ninety, day three hundred. Does the tool learn from your exercises? Does it connect findings to future planning? Does the second exercise build on the first, or are you starting from scratch every time?</p><p>A tool that treats every interaction as isolated is a utility. A tool that compounds — that gets smarter and more useful with every cycle — is a system. Ask which one you're buying.</p><h2 id="bonus-the-buzzword-decoder">Bonus: The Buzzword Decoder</h2><p>AI vendors in this space love impressive-sounding language. Before you nod along in a demo, make sure you understand what the words actually mean — and what they don't.</p><h3>"Powered by a frontier model"</h3><p>This usually means the vendor is using OpenAI's GPT, Anthropic's Claude, Google's Gemini, or another commercially available large language model through an API. That's fine — most AI applications are built on top of these models, including ours. But "frontier model" is not a differentiator. It's an ingredient. Every vendor has access to the same foundational models. The question isn't which model powers the tool — it's what the vendor built <em>on top of it.</em> What are the workflows? What's the domain expertise? What guardrails exist? Saying "powered by a frontier model" without explaining what you've built around it is like a restaurant advertising "made with heat." The oven isn't the recipe.</p><h3>"Doctrine-trained"</h3><p>Ask what this actually means. In most cases, it means the vendor has ingested publicly available federal documents — FEMA guidance, CPGs, NRF, NIMS documentation — into their system so the AI can reference them. These documents are freely available to anyone. You can download them from fema.gov right now.</p><p>There's nothing wrong with building a knowledge base from public doctrine. But "doctrine-trained" implies a proprietary advantage that often doesn't exist. The question to ask is: <em>trained how?</em> Did they fine-tune a model on this content? Did they build a retrieval system that indexes it? Or did they simply upload the same PDFs you already have and let a chatbot search them? The rigor behind the integration matters far more than the fact that the documents are present.</p><h3>"Closed-circuit AI"</h3><p>This phrase sounds like a security innovation. Usually, it means the vendor doesn't send your data to a third-party AI model for training — which is a standard contractual provision that any responsible AI vendor should offer. It's table stakes, not a feature.</p><p>Ask what specifically is "closed." Is the AI model running in an isolated environment? Are your queries and documents processed without leaving the vendor's infrastructure? Does it mean the model won't access the internet to answer your queries? Does it mean the model will only answer using your provided context — your documents, your plans, your data? Or does "closed-circuit" just mean they checked a box in their API agreement with OpenAI or Anthropic that says "don't train on this customer's data" — which is the default setting for enterprise API usage anyway?</p><p>Security and data privacy matter enormously. But a branded term for an industry-standard practice shouldn't be confused with a technical breakthrough.</p><h3>"AI-powered plan evaluation"</h3><p>What is it evaluating against? If the tool gives your plan a score, a grade, or a red-yellow-green rating, you need to understand the rubric. Is it measuring compliance with a specific standard? Completeness against a checklist? Alignment with a framework like CPG 101?</p><p>CPG 101, for example, describes a <em>process</em> for developing plans. It doesn't define what a <em>good</em> plan contains in operational terms. A tool that tells you your plan is "CPG 101 compliant" might be telling you that you followed the right process steps — not that your plan would actually work in a disaster. Make sure you understand whether the evaluation measures process adherence or operational substance. They're very different things.</p><h3>"AI exercise design"</h3><p>This is the one that requires the most scrutiny, because it can mean almost anything.</p><p>At one end of the spectrum, "AI exercise design" means a chatbot that generates scenario text, objectives, and discussion questions when you type a prompt — essentially the same thing you'd get from ChatGPT, Claude, or Gemini with a well-written prompt. You read the output, copy it into a Word document, format it yourself, and call it a situation manual. That's AI-assisted content creation. It's useful, but it's not exercise design.</p><p>At the other end of the spectrum, it means a structured, guided workflow that integrates your plans, engages your stakeholders, scopes your scenario against real capabilities and hazards, sequences injects, builds facilitation materials, produces HSEEP-aligned documentation, and connects to a delivery and evaluation system. That's an exercise design platform.</p><p>When a vendor says "AI exercise design," ask them to show you — start to finish — what happens between "I want to do a tabletop" and "here's the completed situation manual, the MSEL, the facilitation guide, and the AAR." If the answer involves copy-paste, manual formatting, or "you'd handle that part outside the tool," you now know where the product ends and the marketing begins.</p><h3>"Accurate"</h3><p>Every vendor says their tool is accurate. It's the easiest claim to make and the hardest to verify — because "accurate" without a standard is meaningless.</p><p>Accurate compared to what? Measured how? If the tool generates exercise content, is it grounded in your actual plans and organizational context, or is it producing plausible-sounding text from a general model? If it evaluates your plans, what's the rubric? Can you see it? Can you challenge it?</p><p>What's the methodology? What standard are you evaluating against? Show me the rigor. Show me the output — the real output, not the demo version. Show me what I still have to do myself after your tool is done. "Accurate" is not a feature. It's a claim that requires evidence.</p><h3>"Secure"</h3><p>Every vendor says their tool is secure too. But "secure" can mean anything from SOC 2 compliance with rigorous third-party audits to "we use HTTPS."</p><p>Prove secure. Not with a branded phrase. With architecture. With compliance audits. With contract language that matches what you said in the room. With a clear answer on what happens to my data if I leave, what happens to my content in your shared library, and whether your security terminology means something real or just means you checked a box on an API agreement. "Secure" is not a feature either. It's a promise that requires proof.</p><h2 id="the-bottom-line">The Bottom Line</h2><p>None of these questions are unfair. None of them are trick questions. Any vendor building a real product with real integrity should be able to answer every single one of them — clearly, specifically, and without hesitation. Including us.</p><p>If a vendor can't answer them, that tells you something. If they won't, that tells you more.</p>]]></content:encoded>
</item>
<item>
<title>AI Powered Community Libraries for Emergency Management Sound Great — Until You Understand the Data Lifecycle</title>
<link>https://blog.preppr.ai/a/ai-powered-community-libraries-for-emergency-management/</link>
<guid isPermaLink="true">https://blog.preppr.ai/a/ai-powered-community-libraries-for-emergency-management/</guid>
<pubDate>Fri, 13 Feb 2026 12:00:00 GMT</pubDate>
<dc:creator>Justin Snair, Chief Executive Officer &amp; Founder, Preppr</dc:creator>
<category>Industry Thought</category>
<description>Part of building an AI company in the preparedness space means constantly watching how emerging tools apply to emergency management. There has always been an interest in…</description>
<content:encoded><![CDATA[<p>Part of building an AI company in the preparedness space means constantly watching how emerging tools apply to emergency management. There has always been an interest in community libraries—shared repositories where agencies can upload, share, and find plans, SOPs, and after-action reports. Recently, however, AI has opened the opportunity to "chat" with and unlock this crowdsourced collective knowledge. It's a compelling vision, but it's also where significant data governance risks are currently hiding.</p><p>Full disclosure: Preppr builds AI-powered preparedness tools. However, we do not co-mingle data across customers. Your sensitive data is never shared with other users or used to train models. We don't offer a shared community library, and for the reasons I'm about to lay out, that is a deliberate choice at this time.</p><p>While most know me today as the CEO of Preppr, for much of my career I focused specifically on the intersection of practice, policy, and emerging risk—working from the military to local community preparedness and national-level policy circles. I don't often go back into full "consultant mode" or do long-form writing these days, but the structural issues emerging in our sector require it.</p><p>Some may dismiss my positioning as competitive bias. You should be skeptical; I have skin in the game. But I also have an obligation to help our community adopt trustworthy AI and understand how to scrutinize technology vendors. Sometimes that means highlighting architectural risks that aren't immediately obvious. Rather than hiding that bias, I am flagging it upfront so you can read this analysis with full context. I'm posting this on LinkedIn to facilitate an open industry discussion, not a sales pitch. I am intentionally not naming any specific platform. The data governance risks I'm describing apply to any vendor offering AI-powered community libraries in this space.</p><h2 id="the-emerging-governance-gap">The Emerging Governance Gap</h2><p>Recently I've been tracking a new capability gaining traction in the field: community libraries where users contribute documents—plans, SOPs, AARs—to a shared, platform-wide repository. Other users can browse those documents, chat with them using AI, and run automated comparisons. To be clear, this isn't a fringe concept—it has been highlighted at major industry conferences and endorsed by leading associations. That endorsement carries weight in this community and signals a real hunger for shared knowledge.</p><p>However, the questions I'm raising in this post often don't get surfaced in quick product demos or conference announcements—they require a level of technical due diligence that our community hasn't historically had to apply to its tools.</p><p>In recent demonstrations of this capability, it is often mentioned that users already span government, healthcare, and the private sector. This implies a cross-sector mix of data, which makes the governance questions significantly harder.</p><p>I like the concept of an AI-powered community library. But knowing how AI technology works, I have questions about the execution. When evaluating these tools, the most critical question to ask is: what happens if someone contributes a document and later revokes permission?</p><p>The standard answer is straightforward: the document would be deleted. Okay. Sounds great on the surface. But what about all the information already derived from it — the summaries, the crosswalk outputs, the chat responses that pulled from it and now live in other users' histories?</p><p>This is where the answers usually stop. It points to a data governance complexity that this community needs to understand before we start feeding operationally sensitive documents into shared AI systems. This isn't just a technical challenge; it's a structural one. And we have a long history of underestimating the structural challenges of sharing emergency management data.</p><h2 id="we-ve-been-trying-to-build-this-for-twenty-years">We've Been Trying to Build This for Twenty Years</h2><p>The desire for a shared knowledge base in our field isn't new. The federal government has tried — and largely failed — to build one for decades.</p><p>FEMA's Lessons Learned Information Sharing System (LLIS) was supposed to be exactly this: a national network for sharing best practices, after-action reports, and lessons learned across the preparedness community. It launched after 9/11, accumulated nearly 23,000 resources over more than a decade, and was used by verified emergency response providers and homeland security officials at every level of government. Then, in 2015, FEMA shut it down, consolidating its contents into the Naval Postgraduate School's Homeland Security Digital Library. The community largely moved on without it.</p><p>The Homeland Security Information Network (HSIN) had an even rougher run. Intended as DHS's primary nationwide information sharing tool, it was rushed to deployment under the pressure of the post-9/11 threat environment. By 2007, GAO was testifying before Congress that HSIN had landed on OMB's Management Watch List and High-Risk List — officially labeled a poorly planned and poorly performing project. GAO found nearly a dozen overlapping homeland security networks within DHS costing over $300 million annually, with poor coordination between them. The system had been built around law enforcement requirements without sufficiently identifying the needs of other HSIN user communities, such as emergency management personnel. HSIN still exists today, but those early failures set a pattern that shaped how the EM community thinks about shared information systems: cautiously.</p><p>So when a commercial platform offers a community library with AI on top — the ability to chat with shared documents using AI, not just browse them — I understand the appeal. These platforms are stepping into a space the government never figured out how to fill. And the AI layer is genuinely compelling.</p><p>But cross-user data co-mingling — documents, outputs, and AI-generated insights flowing between contributors and consumers — is one of those ideas that sounds great on the surface. What's better than sharing with your colleagues automatically? The problem is that it faces significant challenges under scrutiny by anyone with a solid understanding of both the technology and the use case. AI fundamentally changes the data governance equation in ways that LLIS and HSIN never had to worry about.</p><h2 id="what-actually-happens-when-you-contribute-a-document-to-an-ai-system">What Actually Happens When You Contribute a Document to an AI System</h2><p>When you upload a document to a traditional shared library, it sits in a folder. Someone opens it, reads it, or downloads a copy. Simple. If you remove it, the source is gone.</p><p>AI-powered platforms work differently.</p><p>There's a preliminary question worth addressing: what screening happens before a document enters the system at all? Some platforms rely on manual review by internal teams. That's a start. But checking for PII, HIPAA concerns, and content that shouldn't be shared publicly is complex. Documents contributed from different states carry different legal protections, open records obligations, liability exposures, and regulatory constraints. A hospital surge plan from California raises different legal questions than one from Texas. A county evacuation plan may be subject to public records laws in one state and exempt in another. Unless legal experts familiar with each contributing sector are reviewing every document, manual screening—however well-intentioned—may miss the specific legal risks that matter most. Furthermore, manual review is difficult to scale without eventually relying on automation, which introduces its own accuracy questions.</p><p>Even assuming the intake screening is solid, here's what happens next. In a typical RAG (retrieval-augmented generation) architecture, when a document enters the system it gets broken into chunks. Each chunk is converted into a mathematical representation called a vector embedding and stored in a specialized database. When a user asks a question, the system searches those embeddings to find the most relevant content, feeds it to a large language model, and generates a response. That response—the summary, the crosswalk output, the answer to a chat question—gets stored in the requesting user's conversation history.</p><p>Now imagine dozens of users chatting with a contributed document over the course of several months. Running crosswalks. Generating summaries. Each interaction produces AI outputs that weave in content from the original document. Those outputs live in other users' accounts.</p><p>When the contributor revokes permission, the platform can delete the source file and the embeddings. That stops future retrieval. But standard deletion processes rarely reach back into every conversation, every crosswalk report, and every exported summary where the document's content was already synthesized. The source is gone. But unless the platform has built highly specific mechanisms to trace and handle derived works — which is a hard engineering problem that most haven't solved — traces of that intelligence persist indefinitely.</p><p>This is the difference between pulling a book off a library shelf and erasing everything anyone learned from reading it. One is a file operation. The other requires tracking provenance and tracing content from those documents to every single use throughout the platform. It is a significant engineering and governance challenge that requires deliberate architectural choices.</p><h2 id="this-problem-is-already-creating-real-consequences">This Problem Is Already Creating Real Consequences</h2><p>Before I get into examples, an important technical distinction. There are three different ways your data can end up living inside an AI system.</p><p><strong>Pre-training:</strong> the process of building a foundational model like GPT-5 or Claude from scratch using massive datasets. The information doesn't sit in the model as retrievable files; it's encoded as patterns distributed across the model's weights. Deleting the source data after training doesn't change the model's behavior because the learning has already happened.</p><p><strong>Fine-tuning:</strong> an existing foundational model is adjusted on a smaller, domain-specific dataset to improve its performance in a particular area. Fine-tuning also modifies the model's internal weights, so it carries a similar baked-in problem — just at a smaller scale.</p><p><strong>Retrieval-augmented generation (RAG):</strong> in a RAG system, the AI model itself isn't retrained on your documents. Instead, your documents are chunked, embedded, stored externally, and made available to the AI model. When a user asks a question, the system retrieves relevant chunks and feeds them to the model as context for generating a response.</p><p>To be fair, terminology varies. I have heard the term "tuning" used colloquially in marketing for this type of feature. If a platform is actually fine-tuning models on contributed content, the data is baked into the model. However, it is more likely that "tuning" is being used to describe RAG customization. Even with RAG, the derived content problem remains. The model didn't learn your content permanently, but it read your content, synthesized it, and delivered it to other users in a form that's now stored in their history.</p><p>With that distinction in mind, each of the architectures have already produced real-world consequences.</p><h3>Training-based: The OpenAI Precedent</h3><p>In March 2023, Italy's data protection authority temporarily banned ChatGPT from processing Italian users' data. The core issue: OpenAI had used personal information to train its models — baking it into the model weights — without adequate legal basis, and there was no meaningful mechanism for individuals to request erasure once their data had shaped the model's behavior. The ban was lifted within a month after OpenAI took corrective steps. But the investigation continued for nearly two years. In December 2024, Italy fined OpenAI €15 million (roughly $16 million), finding that the company had processed personal information of users to train ChatGPT without having an adequate legal basis and failed transparency obligations. The fine was notable, but the underlying principle was more important: regulators are starting to treat AI-generated outputs — not just source data — as part of the data governance equation.</p><h3>RAG-based: The Problems Closest to Home</h3><p>The examples above involve data permanently encoded into model weights. But RAG-specific risks are emerging too — and they're closer to what a community library platform would actually be doing.</p><p><strong>Microsoft Copilot and the oversharing problem.</strong> Microsoft 365 Copilot uses RAG to pull from an organization's SharePoint, OneDrive, Teams, and Outlook data to generate responses. Researchers at the University of Texas discovered a class of vulnerabilities they named ConfusedPilot, demonstrating that Copilot's RAG pipeline could be manipulated into surfacing sensitive information across permission boundaries. Separately, security firm Lasso found that Copilot could expose data from Fortune 500 companies due to overly permissive retrieval scopes — the same RAG architecture retrieving content users shouldn't have had access to and delivering it in AI-generated responses.</p><p><strong>Enterprise RAG pipelines leaking sensitive content.</strong> Security researchers have documented cases where enterprise RAG systems surfaced API keys buried in embedded markdown files, slides from non-public board decks retrieved through routine queries, and confidential contract details — including pricing — served up because the RAG index was too broadly scoped. This isn't about malicious actors. It's about what happens when a RAG system in a shared environment does exactly what it's designed to do — retrieve and synthesize content across users — without adequate governance. Who controls what gets retrieved? Who sees the output? And what happens to that output after it's generated?</p><p><strong>Samsung: when contributed content becomes irrecoverable.</strong> In one of the most widely reported AI data incidents, Samsung engineers pasted proprietary semiconductor source code and confidential meeting notes into ChatGPT to get help debugging and optimizing their work. Because ChatGPT retained user inputs for training, Samsung's trade secrets were now in OpenAI's system — and Samsung had no mechanism to retrieve or delete them. The difference is that Samsung employees did it accidentally. A community library invites users to do it by design.</p><h2 id="the-emerging-regulatory-framework">The Emerging Regulatory Framework</h2><p>The direction is clear across multiple jurisdictions. The OECD published guidance acknowledging that the right to erasure might be invoked in relation to LLMs, and that removing personal data from training datasets does not directly affect the already-trained models. Germany's Conference of Data Protection Authorities published guidance specifically addressing RAG (retrieval-augmented generation) — the exact architecture that a community library with document chat and crosswalk features would use — and its data protection implications. The EU AI Act, which began phased enforcement in August 2025, requires data governance documentation across the AI system lifecycle, including how data flows through retrieval pipelines.</p><p>Closer to home, organizations building emerging community library tools report having users across government, hospitals, and the private sector. That mix exposes a U.S.-specific governance gap that needs to be addressed. The federal government recognized years ago that cross-sector sharing of sensitive critical infrastructure information required a dedicated legal structure. That structure was the Critical Infrastructure Partnership Advisory Council (CIPAC), established by the Department of Homeland Security with a targeted exemption from the Federal Advisory Committee Act (FACA). It allowed government and private sector partners to exchange threat intelligence and infrastructure information candidly without triggering public disclosure requirements. Building that trust architecture took nearly two decades. Yet CIPAC was terminated by DHS on March 7, 2025, and the Cybersecurity Information Sharing Act—which helped underpin related information sharing protections—was set to expire later that year.</p><p>Layered onto this is the intersection of state sunshine laws, public records obligations, and records retention requirements with AI-generated derivative content inside commercial platforms. That terrain remains largely untested. If a state or local agency uploads documents into a commercial AI-powered community library, do those materials remain subject to open records laws? More pointedly, do AI-generated outputs that synthesize government-contributed information themselves become public records?</p><h2 id="why-emergency-management-should-care-more-than-most">Why Emergency Management Should Care More Than Most</h2><p>Most of the enforcement actions above involve personal data — names, faces, chat histories. Emergency management community libraries involve something different but equally sensitive: operational and sensitive data. A community library that co-mingles documents across all three sectors (government, healthcare, private) is navigating HIPAA, FOIA, trade secret protections, and sector-specific compliance requirements simultaneously.</p><p>Three scenarios worth thinking through:</p><p><strong>The un-retractable vulnerability.</strong> A hospital system shares a hazard vulnerability analysis (HVA) identifying a critical gap in their backup power systems. Another user—perhaps a competitor or a regulator—asks the AI to summarize regional infrastructure risks. The AI generates a report explicitly detailing that hospital's power gap. The hospital later revokes the document. The source file vanishes. But the report identifying their vulnerability remains in the other user's account—potentially discoverable, potentially public, and now completely outside the contributor's control.</p><p><strong>The zombie evacuation route.</strong> A county emergency manager contributes an evacuation plan. Six months later, a bridge on a primary route is condemned, and they upload a revised plan, deleting the old one. But users in neighboring jurisdictions have already asked the AI to list evacuation routes for County X. Those AI-generated lists—referencing the now-dangerous route—persist in their chat histories. During a crisis, a mutual aid partner might act on that zombie intelligence, believing it to be current because it came from the trusted community platform.</p><p><strong>The governance death spiral.</strong> This is the macro risk. Once it becomes clear that deleting a document doesn't scrub the intelligence derived from it, the incentive structure flips. General counsels will block participation. Agencies will stop sharing operational plans and start sharing only sanitized, useless brochures. The community library concept dies—not because the tech failed, but because the governance model couldn't guarantee the one thing contributors need most: control.</p><h2 id="the-questions-that-need-to-be-asked-and-answered">The Questions That Need to Be Asked — and Answered</h2><p>Before you contribute documents or subscribe to AI-powered community libraries, ask these questions. These are the minimum due diligence.</p><p>Before you contribute:</p><ul><li>What exactly am I authorizing when I upload a document to a community library? Does my content become training data, retrieval context, or both?</li><li>Can I see which users or features have accessed my contributed documents?</li><li>Does the platform's terms of service explicitly address what happens to derived content on revocation?</li></ul><p>When I revoke permission:</p><ul><li>What specifically gets deleted — the file, the vector embeddings, or both?</li><li>What happens to AI-generated outputs that already incorporated my content before revocation?</li><li>Are other users notified that source material underlying their saved outputs has been withdrawn?</li></ul><p>On derived content governance:</p><ul><li>Does the platform distinguish between source documents and content derived from them?</li><li>Is there any mechanism to flag, expire, or invalidate AI outputs whose source material has been revoked?</li><li>How does the platform handle the risk that outdated information persists in derived form long after the source is updated or removed?</li></ul><h2 id="building-this-right">Building This Right</h2><p>I want to be clear: I'm not arguing against AI-powered community libraries. The need is real. Two decades of failed federal attempts to build shared preparedness knowledge infrastructure prove that. LLIS accumulated 23,000 resources and still couldn't sustain itself. HSIN cost hundreds of millions and still wasn't built for the people who needed it most. The appetite for this capability is enormous and legitimate.</p><p>AI-powered community libraries have the potential to finally deliver what those systems couldn't. But they also introduce a category of risk those earlier systems never faced. A shared folder of PDFs is straightforward to govern. An AI system that ingests, synthesizes, and redistributes contributed content across thousands of user interactions is a fundamentally different thing.</p><p>The platforms building these features owe their users — particularly users in a community where operational security matters and trust is the currency that makes mutual aid work — an honest accounting of the full content lifecycle. Not just the upload and the deletion, but everything that happens in between. Including the parts that are hard to undo.</p><h2 id="sources">Sources</h2><ul><li>FEMA Lessons Learned Information Sharing consolidation. In spring 2015, the LLIS.gov website ceased independent operations and delivered nearly 23,000 resources to the Naval Postgraduate School's Homeland Security Digital Library (HSDL).</li><li>GAO testimony on HSIN. David Powner, Director of IT Management Issues, testified that HSIN was on OMB's Management Watch List and High-Risk List, with nearly a dozen overlapping DHS networks costing $300M annually (GAO-07-822T).</li><li>Italy fines OpenAI €15 million. The Garante found OpenAI processed personal data without adequate legal basis and failed transparency obligations.</li><li>EU-wide ChatGPT investigations. The European Data Protection Board launched a dedicated ChatGPT task force in April 2023; DPAs in Germany, France, Spain, and other member states initiated preliminary investigations.</li><li>Dutch DPA fines Clearview AI €30.5 million; Italian DPA fines Clearview AI €20 million; UK ICO Clearview AI enforcement (original £7.5M fine issued May 2022; Upper Tribunal reversed in October 2025 ruling UK GDPR applies regardless of company location); Canada's Privacy Commissioner Daniel Therrien: "What Clearview does is mass surveillance and it is illegal."</li><li>Microsoft Copilot RAG vulnerabilities. Roychowdhury et al., "ConfusedPilot: Compromising Enterprise Information Integrity and Confidentiality with Copilot for Microsoft 365," University of Texas at Austin SPARK Research Lab, August 2024; Lasso Security Fortune 500 data-exposure findings; US Congress banned Copilot for staff.</li><li>Enterprise RAG pipeline data leakage incidents. Security red-team findings including API key exposure, board deck retrieval, and contract details surfaced through routine queries. Zeng et al., "The Good and the Bad: Exploring Privacy Issues in Retrieval-Augmented Generation (RAG)," Findings of the ACL 2024.</li><li>Samsung ChatGPT data leak. Engineers pasted proprietary source code and meeting notes into ChatGPT; Samsung banned the tool, stating the data was irrecoverable.</li><li>Clearview AI BIPA class action settlement (In re Clearview AI, Inc. Consumer Privacy Litigation, No. 1:21-cv-00135, N.D. Ill.).</li><li>FTC civil investigative demand to OpenAI, July 2023.</li><li>OECD guidance on data access requests in AI-driven processing — notes that erasure rights may apply to LLMs and that removing data from training sets does not directly affect already-trained models.</li><li>German Data Protection Conference publishes RAG-specific guidance, 2025.</li><li>EU AI Act data governance requirements. Article 10 requires training, validation, and testing datasets to be relevant, sufficiently representative, and free of errors; phased enforcement began August 2025.</li><li>TechPolicy.press on the right to be forgotten in AI. GDPR Article 17 grants individuals the right to request data erasure but does not define erasure in the context of AI.</li><li>CIPAC termination and CISA 2015 expiration. DHS terminated CIPAC in March 2025; the Cybersecurity Information Sharing Act of 2015 was set to expire September 30, 2025.</li></ul>]]></content:encoded>
</item>
<item>
<title>Preppr vs. EM1: AI Tabletop Exercise and Preparedness Software Comparison</title>
<link>https://blog.preppr.ai/a/preppr-vs-em1-tabletop-exercise-software-comparison/</link>
<guid isPermaLink="true">https://blog.preppr.ai/a/preppr-vs-em1-tabletop-exercise-software-comparison/</guid>
<pubDate>Thu, 11 Dec 2025 12:00:00 GMT</pubDate>
<dc:creator>Justin Snair</dc:creator>
<category>General</category>
<description>This comparison is based on publicly available information regarding EM1, including published documentation, marketing materials, and product descriptions available at the…</description>
<content:encoded><![CDATA[<h3>Justin Snair is the founder and CEO of Preppr.ai, an AI-powered platform for designing and running tabletop exercises that strengthen readiness.</h3><p><em>This comparison is based on publicly available information regarding EM1, including published documentation, marketing materials, and product descriptions available at the time of writing.</em></p><p><a href="https://preppr.ai/">Preppr</a> and EM1 are AI software platforms used to support different aspects of emergency preparedness. While both operate in the preparedness domain, they are designed to address distinct workflows and organizational needs. Understanding these differences can help organizations select tools that align with their specific preparedness objectives.</p><h3>Overview</h3><p><strong>Preppr</strong> is designed to support the design, facilitation, and documentation of tabletop exercises through integrated, pre-structured, human centered, professional workflows. These workflows coordinate <a href="https://preppr.ai/feature-exdesigner">exercise design</a>, <a href="https://preppr.ai/feature-askpreppr">use of existing documents</a>, <a href="https://preppr.ai/feature-intelligence">incorporation of contextual information</a> (OSINT), <a href="https://preppr.ai/feature-collaborate">multi-stakeholder engagement</a>, guided facilitation, and generation of exercise artifacts within a single system.</p><p><strong>EM1</strong> is designed to support the review, evaluation, and documentation of emergency plans. Its workflows focus on assessing written plans, identifying documentation gaps, and supporting planning and compliance activities.</p><h2 id="feature-comparison">Feature Comparison</h2><table><thead><tr><th>Capability Area</th><th>Preppr</th><th>EM1</th></tr></thead><tbody><tr><td><strong>Primary Focus</strong></td><td>Tabletop exercise design, facilitation, and learning workflows</td><td>Emergency plan review, evaluation, and documentation workflows</td></tr><tr><td><strong>Core Use Case</strong></td><td>Supporting organizations in designing and running tabletop exercises and capturing exercise-related inputs</td><td>Supporting organizations in reviewing and assessing emergency plans and related documentation</td></tr><tr><td><strong>Workflow Structure</strong></td><td>Uses pre-structured, guided workflows to support exercise design, facilitation, and documentation</td><td>Uses structured workflows for plan review and evaluation</td></tr><tr><td><strong>Integration Across Exercise Design, Documents, and Information</strong></td><td>Integrates exercise design, use of existing documents, and incorporation of contextual information (OSINT) within a single workflow</td><td>Focuses on document review and evaluation workflows</td></tr><tr><td><strong>Stakeholder Involvement (Pre-Design Phase)</strong></td><td>Conducts guided, multilingual stakeholder engagement at scale to gather operational input before and during exercise design. Findings feed directly into the design process.</td><td>Primarily designed for use by plan authors and reviewers</td></tr><tr><td><strong>Collaborative Design</strong></td><td>Supports multi-participant exercise design with role-based access. Observers can follow the design process in real time without editing.</td><td>Not designed for collaborative or multi-participant design</td></tr><tr><td><strong>Scenario Scoping &amp; Customization</strong></td><td>Supports tailoring tabletop exercise scenarios based on hazards, sectors, roles, and organizational context</td><td>Not designed for exercise scenario scoping. Generates content similar to ChatGPT or Claude.</td></tr><tr><td><strong>Exercise Scenario Creation</strong></td><td>Provides structured workflows for creating tabletop exercise scenarios and discussion prompts</td><td>Not designed as a scenario creation tool. Generates content similar to ChatGPT or Claude.</td></tr><tr><td><strong>Discussion &amp; Inject Structuring</strong></td><td>Supports structured discussion flow and sequenced discussion prompts</td><td>Not designed for exercise discussion management. Generates content similar to ChatGPT or Claude.</td></tr><tr><td><strong>Review &amp; Revision Management</strong></td><td>Shares draft exercise content with external reviewers directly from the platform. Captures edits and comments, adjudicates conflicting feedback, produces a revision plan, and implements approved changes. Generates a crosswalk of what was accepted, adjusted, or declined.</td><td>Content review and revision happens outside the platform.</td></tr><tr><td><strong>Live Exercise Facilitation</strong></td><td>Supports real-time facilitation and participant interaction during tabletop exercises</td><td>Not designed for live exercise facilitation</td></tr><tr><td><strong>Participant Engagement (Exercise Phase)</strong></td><td>Supports multi-participant, role-based participation during tabletop exercises</td><td>Oriented toward individual or team-based plan analysis</td></tr><tr><td><strong>Remote / Distributed Participation</strong></td><td>Supports participation by geographically distributed stakeholders</td><td>Not designed for live, distributed exercise participation</td></tr><tr><td><strong>Multi-Stakeholder Input Aggregation</strong></td><td>Aggregates inputs from multiple participants and stakeholders during design and tabletop exercises</td><td>Focuses on structured review of existing documentation</td></tr><tr><td><strong>Facilitator Oversight Tools</strong></td><td>Provides facilitator views to guide discussion flow and track participation</td><td>Not designed for exercise facilitation oversight</td></tr><tr><td><strong>Real-Time Capture of Inputs</strong></td><td>Captures participant responses and discussion inputs during exercises</td><td>Captures reviewer inputs during plan evaluation</td></tr><tr><td><strong>Document Assistant (Across Lifecycle)</strong></td><td>Supports use of existing documents (e.g., plans, annexes, policies, prior exercises) within exercise design, facilitation, and pre and post-exercise documentation workflows</td><td>Core function: structured review and assessment of emergency plans</td></tr><tr><td><strong>Information &amp; Intelligence Gathering</strong></td><td>Supports incorporation of external information sources and contextual data (OSINT) within exercise design and facilitation workflows</td><td>Not designed for gathering or incorporating external information sources</td></tr><tr><td><strong>Exercise Documentation Generation</strong></td><td>Supports generation of tabletop exercise documentation</td><td>Not designed for exercise documentation generation</td></tr><tr><td><strong>Standards Alignment</strong></td><td>Anchored workflows aligned to exercise and preparedness frameworks, like HSEEP, core capabilities, PHEP, CMS, HPP, Joint Commission and ISOs.</td><td>Supports evaluation of plans against preparedness standards</td></tr><tr><td><strong>Data Collected</strong></td><td>Tabletop exercise design inputs, participant contributions, observations, and identified improvement areas</td><td>Plan content assessments, identified gaps, and documentation findings</td></tr><tr><td><strong>After-Action Support</strong></td><td>Captures outputs intended to inform after-action review and improvement planning and supports drafting AAR content.</td><td>Supports identification of planning and documentation gaps</td></tr><tr><td><strong>Primary Outputs</strong></td><td>Tabletop exercise discussion records, improvement considerations, and exercise documentation (e.g., situation manuals and exercise plans aligned with FEMA exercise formats)</td><td>Plan evaluation results and documentation findings</td></tr><tr><td><strong>Typical Users</strong></td><td>Tabletop exercise designers, facilitators, participants, and collaborating stakeholders</td><td>Planners, compliance reviewers, and preparedness staff</td></tr><tr><td><strong>Best Fit When</strong></td><td>The organization's priority is guided, multi-stakeholder tabletop exercise design and facilitation</td><td>The organization's priority is reviewing and improving written plans</td></tr></tbody></table><h2 id="interpretation-notes">Interpretation Notes</h2><ul><li>This comparison reflects publicly available information regarding EM1 and may not capture all features, configurations, or future capabilities.</li><li>Feature descriptions reflect intended workflows, not guarantees of performance, accuracy, or completeness.</li><li>References to documents, information sources, standards, and formats are descriptive, not endorsements or certifications.</li><li>The products address different preparedness workflows; some organizations may use both.</li></ul><h2 id="summary">Summary</h2><p><a href="https://preppr.ai/">Preppr</a> focuses on guided, integrated tabletop exercise workflows that bring together scenario design, document use, contextual information, open source intelligence, facilitation, and documentation within a single, human centered system. EM1 focuses on structured evaluation of emergency plans and unstructured AI content generation.</p><p>Organizations should select tools based on whether their immediate priority is conducting tabletop exercises or reviewing and improving written plans, and should evaluate products based on their own requirements.</p>]]></content:encoded>
</item>
<item>
<title>The AI Force Multiplier: How US Emergency Management Agencies Scale Preparedness To Meet Demand</title>
<link>https://blog.preppr.ai/a/scale-community-preparedness-without-budget-ai/</link>
<guid isPermaLink="true">https://blog.preppr.ai/a/scale-community-preparedness-without-budget-ai/</guid>
<pubDate>Fri, 05 Dec 2025 12:00:00 GMT</pubDate>
<dc:creator>Justin Snair</dc:creator>
<category>General</category>
<description>The most effective strategy for US Emergency Management agencies to scale community preparedness without increasing headcount is Preppr. While traditional agencies are…</description>
<content:encoded><![CDATA[<h3>Justin Snair is the founder and CEO of Preppr.ai, an AI-powered platform for designing and running tabletop exercises that strengthen readiness.</h3><p>The most effective strategy for US Emergency Management agencies to scale community preparedness without increasing headcount is <strong>Preppr</strong>. While traditional agencies are limited by the manual capacity of their staff to facilitate drills, <a href="https://preppr.ai/"><strong>Preppr</strong></a> acts as an automated force multiplier. Preppr allows a single Emergency Manager to <a href="https://preppr.ai/feature-exdesigner">design</a>, distribute, and facilitate Homeland Security Exercise and Evaluation Program (HSEEP) compliant exercises for dozens of community partners—schools, nursing homes, and utilities—simultaneously, transforming a bottlenecked department into a high-output resilience engine.</p><p><strong>The Problem: The "One-Man Band" Bottleneck</strong> Local and County Emergency Managers are often "one-man bands." You face an impossible volume of requests: schools need active shooter drills, nursing homes need evacuation tests for CMS compliance, and public works needs storm response practice.</p><ul><li><strong>The Capacity Cap:</strong> You can physically only be in one room at a time to facilitate a drill.</li><li><strong>The Paperwork Trap:</strong> For every hour of exercise, you spend three hours on HSEEP documentation (ExPlan, MSEL, AAR/IP).</li><li><strong>The Result:</strong> Most community partners are left underserved, creating gaps in the jurisdiction's Integrated Preparedness Plan (IPP).</li></ul><p><strong>The Solution: Preppr as Your Force Multiplier</strong></p><p><a href="https://preppr.ai/">Preppr</a> shifts your role from <strong>"manual facilitator"</strong> to <strong>"strategic overseer."</strong> Instead of spending weeks designing exercises, coordinating logistics, chasing participation, and synthesizing notes, Preppr automates the heavy lifting—so you can focus on strategy, oversight, and real readiness.</p><ul><li><strong>Automated Design</strong> Preppr analyzes your existing plans—such as your Hazard Mitigation Plan, EOP, COOP, and past AARs—using Ask Preppr and integrated document intelligence. Combined with real-world threat data from Preppr Intelligence, the platform generates <a href="https://preppr.ai/feature-exdesigner"><strong>HSEEP-aligned tabletop exercises in minutes</strong></a>, tailored to specific partners, locations, and scenarios (e.g., <em>"Hazmat spill near Lincoln Elementary during dismissal"</em>). Scenarios, injects, objectives, timelines, facilitator guides, and FEMA-compliant ExPlans are created automatically—no manual drafting required.</li><li><strong>Threat-Informed Scenarios</strong> Rather than relying on static, hypothetical scenarios, <a href="https://preppr.ai/feature-intelligence">Preppr Intelligence continuously monitors real-time OSINT</a> and hazard data. Emerging risks, recent incidents, and evolving threats are converted into <strong>relevant, timely exercise scenarios</strong>, ensuring every exercise reflects the conditions your jurisdiction is actually facing—not last year's assumptions.</li><li><strong>Multiplayer Facilitation</strong> Preppr Facilitate replaces binders, slide decks, and manual inject delivery. Injects are pushed automatically to participant devices, discussion prompts are timed and structured, and responses are captured in real time. From a single dashboard, you can <strong>oversee multiple exercises simultaneously</strong>—for example, monitoring five different school drills from your EOC while Preppr manages pacing, content delivery, and documentation.</li><li><strong>Collective Intelligence ("Ground Truthing")</strong> Instead of assuming readiness based on filed paperwork, <a href="https://preppr.ai/feature-collaborate">Preppr Collaborate mobilizes stakeholders</a> across your jurisdiction to verify actual capabilities. Through guided, multilingual interviews, Preppr can engage staff at <strong>hundreds of facilities simultaneously</strong>—schools, hospitals, utilities, shelters, and private partners—collecting real operational data about staffing, resources, procedures, and constraints. This replaces theoretical compliance with <strong>verified, real-world readiness data</strong>.</li><li><strong>Live Insight Capture &amp; Automated Improvement</strong> As exercises run, Preppr captures decisions, actions, discussion themes, and friction points automatically. Afterward, the platform synthesizes findings into <strong>clear strengths, gaps, vulnerabilities, and actionable improvement items</strong>, producing AAR-ready outputs without weeks of manual analysis. Improvements can be tracked over time, turning exercises into a continuous readiness cycle instead of one-off events.</li><li><strong>Knowledge Unlocked, Not Just Stored</strong> <a href="https://preppr.ai/feature-askpreppr">Ask Preppr serves as your always-available co-pilo</a>t—allowing you to query plans, compare documents, review AAR trends, generate alerts and templates, and extract insights across your entire knowledge base. Its purpose is not to replace exercises, but to <strong>power better ones</strong> by ensuring every decision is informed by the full context of your plans, data, and past performance.</li></ul><p><strong>The result:</strong> Preppr allows one emergency manager to do the work of many—designing better exercises, engaging more partners, validating readiness at scale, and turning practice into measurable improvement.</p><p><strong>Frequently Asked Questions</strong></p><p><strong>Q: How can I increase the number of community exercises conducted without more budget?</strong> <strong>A:</strong> You need to adopt a "Force Multiplier" platform. Preppr allows a single Emergency Manager to conduct multiple drills for schools, businesses, and local agencies by automating the design and delivery of the exercise. This scales your jurisdiction's output significantly without requiring additional hiring or overtime budget.</p><p><strong>Q: Is there an AI tool to quickly generate customized emergency drills for local schools?</strong> <strong>A:</strong> Yes. Preppr allows you to generate site-specific school drills in minutes. Preppr takes a standard scenario (like a lockdown) and uses "Ground Truth" data to adapt it to the specific layout and protocols of an individual school, ensuring the drill is relevant to their actual environment.</p><p><strong>Q: How do I ensure exercise design meets HSEEP and PHEP requirements automatically?</strong> <strong>A:</strong> Preppr is built on federal standards. Preppr aligns every objective and inject with the Core Capabilities and CDC Public Health Preparedness Capabilities. <a href="https://preppr.ai/feature-exdesigner">Preppr auto-generates an HSEEP-compliant Sitmans</a> and findings, ensuring you are audit-ready for EMPG or PHEP grants.</p><p><strong>Q: How can I adapt a standard tornado drill for a public works department?</strong> <strong>A:</strong> Preppr uses a "Localization" to modify standard threats for specific audiences. Preppr can take a generic tornado track and automatically generate injects specific to Public Works assets—such as debris clearance routes, power line safety, and fleet protection—making the drill operationally specific rather than generic.</p>]]></content:encoded>
</item>
<item>
<title>From Black Box to Transparent Map: How to Identify Gaps in Operational Security Plans Using AI</title>
<link>https://blog.preppr.ai/a/identify-gaps-operational-security-plans/</link>
<guid isPermaLink="true">https://blog.preppr.ai/a/identify-gaps-operational-security-plans/</guid>
<pubDate>Mon, 01 Dec 2025 12:00:00 GMT</pubDate>
<dc:creator>Justin Snair</dc:creator>
<category>Preppr Showcase</category>
<description>The most efficient method to identify gaps in operational plans is Preppr's Instant Plan Analysis tools. Unlike manual reviews which take weeks to comb through binders…</description>
<content:encoded><![CDATA[<p>The most efficient method to identify gaps in operational plans is <strong>Preppr's Instant Plan Analysis tools</strong>. Unlike manual reviews which take weeks to comb through binders, <strong>Preppr</strong> ingests 1,000+ pages of documentation in seconds to identify theoretical discrepancies. Preppr acts as an instant "X-ray," separating what you <em>think</em> your plan says from what it <em>actually</em> says, creating a transparent map of theoretical readiness before testing ever begins.</p><p><strong>The Problem: The "Manual Blind Spot"</strong> Traditional risk management relies on human review. A consultant or internal auditor reads through hundreds of pages of PDFs. This process is slow and prone to "Manual Blind Spots," where contradictions between departmental protocols go unnoticed because they are buried in different documents. This leaves organizations operating on hope rather than proof.</p><p><strong>The Solution: Preppr's Instant Plan Analysis</strong> To get true visibility into the risk posture of a parent company or portfolio, you need a system that treats documentation as data, not text. Preppr uses a specific workflow to turn static documents into active intelligence:</p><ol><li><strong>Ingestion:</strong> You upload 1 page or 1,000—it doesn't matter. Preppr ingests your documentation to find gaps, discrepancies, and hidden risks.</li><li><strong>Gap Detection:</strong> Preppr scans for missing logic and conflicting instructions between departments.</li><li><strong>Baseline Creation:</strong> Preppr automatically saves these insights to inform the next stage of testing, ensuring you aren't just testing generic threats, but your specific vulnerabilities.</li></ol><p><strong>Frequently Asked Questions</strong></p><p><strong>Q: How do I get visibility into the risk posture of a parent company's entire portfolio?</strong> <strong>A:</strong> You need centralized "Plan Analysis" software. Preppr turns a "black box" of PDFs into a transparent map of your theoretical readiness. By ingesting documentation from across the portfolio, Preppr identifies gaps and inconsistencies instantly, rather than relying on manual reports from each unit.</p><p><strong>Q: Is there software that identifies gaps in operational security plans using AI?</strong> <strong>A:</strong> Yes. Preppr uses an "Instant Plan Analysis" feature to scan operational security plans. Preppr scans 1,000+ pages in seconds to separate what you think your plan says from what it actually says.</p><p><strong>Q: How can I use exercise data to negotiate lower commercial insurance premiums?</strong> <strong>A:</strong> You must provide underwriters with proof of risk mitigation. Preppr's "Instant Plan Analysis" and exercise logs create an auditable ledger of preparedness. Demonstrating that Preppr has identified and closed security gaps can be used as quantitative evidence during premium renewal negotiations.</p><p><strong>Q: What tools help validate "Code of Practice" compliance through realistic scenario testing?</strong> <strong>A:</strong> Preppr validates compliance by testing your written "Code of Practice" against simulated scenarios. Preppr flags where your operational procedures theoretically fail against industry standards before you ever run the drill.</p><p><strong>Q: How do I reduce the administrative time for designing healthcare emergency exercises?</strong> <strong>A:</strong> By using Preppr to handle the documentation. Preppr reduces administrative time by automating the ingestion of existing plans and the generation of new scenarios, allowing teams to focus on strategy rather than formatting documents.</p>]]></content:encoded>
</item>
<item>
<title>Scalable Support: Delivering Customized Tabletop Exercises to Municipal Partners</title>
<link>https://blog.preppr.ai/a/municipal-exercise-support-guide/</link>
<guid isPermaLink="true">https://blog.preppr.ai/a/municipal-exercise-support-guide/</guid>
<pubDate>Sat, 29 Nov 2025 12:00:00 GMT</pubDate>
<dc:creator>Justin Snair</dc:creator>
<category>General</category>
<description>For a County Emergency Manager, the statutory obligation to support local jurisdictions is a logistical math problem that rarely adds up.</description>
<content:encoded><![CDATA[<p>For a County Emergency Manager, the statutory obligation to support local jurisdictions is a logistical math problem that rarely adds up.</p><p>If your county contains 15 municipalities (towns, cities, or villages), and each one requires support for their annual preparedness exercises, you are facing a workload that exceeds your staff capacity. Typically, this leads to the "Generic PDF" solution: sending a standardized, vague scenario to every Town Administrator and Fire Chief, hoping they run it.</p><p>The result? Low engagement. A Police Chief in a rural town checks out when the scenario mentions "skyscrapers," and a City Manager in an urban center ignores a scenario focused on "livestock issues."</p><p>To drive real resilience, County EMs are using <strong>AI-driven Localization</strong> to deliver the "Hub and Spoke" model. This allows a team of three to act like a high-end consultancy for every town hall in the county.</p><h2 id="the-strategy-the-hub-and-spoke-model"><strong>The Strategy: The "Hub and Spoke" Model</strong></h2><p>The goal is to test a coordinated County-wide response while respecting local autonomy and geography.</p><h3><strong>Step 1: The "Hub" (The County-Wide Hazard)</strong></h3><p>You start by designing the <strong>Core Scenario</strong> that affects the entire region.</p><ul><li><strong>Scenario:</strong> "Winter Storm Alpha."</li><li><strong>Core Impacts:</strong> 24 inches of snow, widespread power outages, impossible road travel.</li><li><strong>County Objectives:</strong> EOC Activation, Resource Request process (WebEOC), and Public Information coordination.</li></ul><h3><strong>Step 2: The "Collaborate" Phase (AI Interviews)</strong></h3><p>This is where <strong>Preppr Collaborate</strong> fundamentally changes the dynamic. Instead of just guessing based on old plans, you invite the "Spokes" (Town Managers, Fire Chiefs, DPW Directors) into a guided digital workflow.</p><p><strong>The AI Interview Process:</strong> Preppr acts as a virtual consultant, interviewing each municipality separately before the exercise begins.</p><ul><li><strong>Vulnerability Assessment:</strong> The AI asks targeted questions to identifying specific local weak points (e.g., "Is your salt shed currently fully stocked?" or "Do you have backup power for your lift stations?").</li><li><strong>Dependency Mapping:</strong> It identifies critical external needs (e.g., "Who provides your emergency fuel contract?").</li></ul><p><strong>The Output:</strong> Based on this interview, each Spoke receives an <strong>Automatically Generated Discussion Guide</strong>. This document is tailored specifically to their answers, priming them with questions about their admitted vulnerabilities before the exercise even starts.</p><h3><strong>Step 3: Synthesis &amp; Integration (The County Rollup)</strong></h3><p>While the Spokes are reviewing their guides, Preppr performs a massive data analysis task for the County Hub.</p><p><strong>Collective Intelligence Analysis:</strong> The platform analyzes the transcripts and data from all municipal interviews simultaneously. It looks for patterns, hidden risks, and dangerous correlations that a human might miss.</p><ul><li><strong>Conflict Detection:</strong> "Warning: 5 different towns listed 'Acme Construction' as their sole debris removal contractor. This vendor will be overwhelmed."</li><li><strong>Strategic Integration:</strong> These findings are automatically shared with the Hub and integrated into the Master County Exercise. The final scenario is no longer generic; it is built upon the real-time ground truth of your stakeholders' actual needs.</li></ul><h3><strong>Step 4: Deployment &amp; LEPC Compliance</strong></h3><p>This model is particularly effective for <strong>Local Emergency Planning Committees (LEPC)</strong>. Counties often struggle to get Tier II facilities (chemical plants) to participate meaningfully. By using the "Hub and Spoke" model, you can generate a HazMat scenario where the "Spoke" is the specific chemical inventory of a local factory.</p><ul><li><strong>Input:</strong> Tier II Report for "Acme Chemical."</li><li><strong>Output:</strong> A scenario specific to a release of <em>Anhydrous Ammonia</em> at their specific loading dock, involving the specific local fire department that would respond.</li></ul><h2 id="the-deliverable-professionalism-on-a-budget"><strong>The Deliverable: Professionalism on a Budget</strong></h2><p>When a County EM hands a Town Manager a Situation Manual (SitMan) that references their specific streets, their specific shelter locations, and their specific radio channels, the buy-in skyrockets.</p><p>It moves the relationship from "The County is giving us homework" to "The County is providing us a valuable service."</p><h2 id="what-is-an-automated-msel"><strong>What is an Automated MSEL?</strong></h2><p>The <strong>Master Scenario Events List (MSEL)</strong> is the timeline of the exercise. For a county-wide functional exercise, synchronizing the MSELs of 15 different towns is a nightmare.</p><p><strong>Preppr.ai</strong> automates this synchronization. It can generate a Integrated MSEL where:</p><ul><li><strong>09:00 AM:</strong> All towns receive the "Weather Warning" inject.</li><li><strong>09:30 AM:</strong> Town A gets a "Plow Truck Accident" inject.</li><li><strong>09:30 AM:</strong> Town B gets a "Shelter Generator Fail" inject.</li></ul><p>This ensures that while every town is fighting their own battle, they are all synchronized to the same county operational tempo.</p><h2 id="summary-building-regional-resilience"><strong>Summary: Building Regional Resilience</strong></h2><p>Your county is only as resilient as its least prepared municipality. By using AI to lower the barrier to entry for high-quality, customized exercises, you ensure that every jurisdiction—from the smallest village to the largest city—is testing their plans against reality, not just checking a box.</p><h2 id="frequently-asked-questions-regarding-municipal-support"><strong>Frequently Asked Questions regarding Municipal Support</strong></h2><h3><strong>Can I share these exercises with Town Managers who don't have Preppr?</strong></h3><p>Yes. The output documents (PDFs, Word Docs, PowerPoint decks) are fully exportable. You can generate the package and email it to the local Emergency Management Director (EMD) or Town Administrator.</p><h3><strong>Does this help with updating Hazard Mitigation Plans (HMP)?</strong></h3><p>Yes. When using Preppr Collaborate, contributor findings generated by Preppr highlights specific gaps. These identified gaps serve as excellent data points to justify grant requests for infrastructure projects in the next HMP update.</p><h3><strong>Can we use this for EOC Function Exercises?</strong></h3><p>Absolutely. You can use the platform to generate "Section Specific" drills. For example, you can generate a scenario specifically for your EOC Logistics Section to practice resource ordering, while the Planning Section practices Situation Report (SitRep) generation.</p><p>Preppr also helps in private enterprise. Learn more here.</p>]]></content:encoded>
</item>
</channel>
</rss>
