Somebody Should Be Recording This

My Operations Chief spent a good chunk of our monthly operations meeting yesterday going through maps. Nothing was broken. Nobody filed a complaint. He’s got about forty years of understanding of why our districts look the way they do, and he’s decided, on purpose, that it needs to get out of his head while he’s still here to explain it. He put it on the agenda twice, back to back, because he knew one slot wasn’t going to cover it. So we put up overlays. Dispatch map against jurisdictional boundaries against annexation history, one on top of another, and he walked us through where they disagree and why they disagree. It was a good meeting. He was doing exactly what a leader is supposed to do.

That meeting is the best tool we own.

What Was Actually in That Room

Here’s a sample of what came out of that hour. Pay attention to what kind of knowledge it is.

There’s a fence post in a field. That fence post marks where our district ends and the next fire district begins. Not a road, not a parcel line, not anything a GIS layer will ever show you. A fence post, in a field, agreed to a long time ago by people who are mostly gone now, and everybody since has just known.

The dispatch map is not completely accurate in certain areas, and he can tell you which ones. Some of that traces to annexations the overlay never caught up with. Which means the system that assigns units to addresses is, in specific and knowable places, wrong.

There’s parks district property sitting inside another fire district’s boundary. Not our jurisdiction. We go anyway, because we protect that parks district’s property, and that’s not a favor or a habit. It’s the correct answer.

None of that is trivia. That’s legal jurisdiction, unit response, and how we tactically operate on arrival. And he didn’t get it by hanging around long enough for it to soak in. He got it by reading, by research, and by sitting through two decades of meetings where these things were actually decided. It’s assembled knowledge. Somebody did the work.

The Map Is Wrong and Nothing Knows It

Sit with the dispatch map part, because it’s the sharpest thing that was in that room.

We have a system of record. It assigns units when the tones drop. It’s authoritative in the only way that counts, which is that it’s what actually happens. And in specific places it’s wrong, and the correction exists, and the correction is stored in a human being who retires eventually.

Now go looking for where that correction is supposed to live. The CAD has no field for “this boundary is wrong and here’s why.” The GIS layer has no slot for “the annexation went through years ago and the overlay never caught up.” The preplan has nowhere to put “we respond here even though it isn’t ours, because of the parks district.” Every system we own is built on the assumption that its own data is correct. Not one of them has a place for the truth about where it’s wrong.

That’s not a bug in any of those products. That’s a category we never built.

The Conversation Already Happened

Now notice the part that should bother us most, because it isn’t the part we usually complain about.

The transfer worked. It happened yesterday, for an hour, in a room. Everything I just told you got said out loud by a man who knows it, to a group of people who needed it. The knowledge moved.

The audience was however many of us were sitting there.

We talk about knowledge capture like the hard part is getting people to give it up. It isn’t. He wanted to give it up. He put it on the agenda twice to give it up, and he’d have put it on again next month if we’d run out of time. Forty years, rationed out in agenda slots. That’s not a knock on him; he’s using the only instrument the job ever handed him. Something north of a quarter of public sector agencies have no clear process for knowledge transfer when somebody walks out the door (NEOGOV, https://www.prweb.com/releases/new-neogov-report-finds-public-sector-ai-adoption-is-growing-but-workforce-readiness-is-lagging-302782940.html), and we’re not one of the ones with a process. We’re one of the ones with a chief who decided to try anyway.

And when the meeting ended, so did the knowledge, except for whatever each of us happened to retain. What I retained is a fence post and a nagging sense that I should ask him about the annexations again before I forget which ones he said.

This is why every fix so far has failed. The instinct is always to add a field. A lessons-learned box on the after-action. A form. A required narrative. It dies every time, and it dies because we aimed at the wrong problem. He was never withholding. There was just nowhere to put it that wasn’t a person’s memory or a document nobody opens.

Here’s what’s different now. The tools showing up in this service can already take how people actually talk and turn it into something structured. That’s exactly what the report writing tools are proving out, even though they’re pointed at the wrong target. That hour yesterday, the overlays and the fence post and the annexation history and the parks district agreement, is precisely the kind of thing that capability could hold onto. Point it at the meeting instead of the PCR and you’re not saving somebody eight minutes of charting. You’re building a memory.

We’re Early, and That’s the Whole Advantage

Nobody’s built this yet. Not us, not a vendor, not anybody. That’s not a gap in the market. That’s a head start, and I’d like us to use it.

Federal agencies are a couple of years ahead of the fire service on AI, and they’re finding out something worth having for free. When a system spends years learning an organization, the understanding it builds doesn’t move. The records export fine. The learning stays behind, welded to whichever platform it grew inside, and there’s no standard for moving it because nobody has even defined what moving it would mean (Federal News Network, https://federalnewsnetwork.com/commentary/2026/06/the-coming-ai-reckoning-slouching-toward-vendor-lock/). They’re learning that on the back end, after signing. We get to learn it before we build.

Think about why your email works. A message from a Gmail account lands in an Outlook inbox because SMTP became an open standard in 1982, before any email company got big enough to stop it. One decision, made early, by people thinking a decade out. It has outlived every business that ever tried to own email. Office documents went the other direction, fought over later against entrenched vendors, and took roughly twenty years to half-work. Same technology curve, opposite ending, and the only variable was whether the standard showed up before or after the market consolidated.

The fire service is standing in 1982 on this one. The memory layer doesn’t exist yet, which means its shape is still ours to define, and defining things is something this field is historically good at. We wrote the NFPA standards. We built NIMS. We made NEMSIS out of nothing because a group of people decided EMS data ought to be comparable across state lines, and nobody handed us that. There’s no reason we can’t decide what a department’s operational memory is, what format it lives in, and who owns it, before it arrives as a checkbox inside somebody’s platform.

Because it’s coming either way. And whatever gets built first becomes the standard by default.

Own the Memory. Rent the Machine.

Which lands on a doctrine short enough to say in one breath.

The corpus is the asset. The engine is a tenant. The after-actions, the near-misses, the annexation history, the reason for the fence post: that lives somewhere open and inspectable that the department holds, in a form we could hand to anybody. Then point whatever AI is best this year at it. When a better one shows up next year, and it will, point that one at it instead. The models are getting better and cheaper and more interchangeable every year, which makes them the wrong thing to build a future on and the right thing to rent. Nobody else can build forty years of our districts. That’s the part with our name on it.

Ten years out, somebody keys up and asks why we’re crossing a district line for a parks property. There are two versions of that moment. In one, a young officer already knows, because the department told him. In the other, everybody shrugs and makes a decision without the one thing that would have made it right.

The department never forgets. We say that like it’s a value we hold. It’s about a decade from being a technical description, and that fence post is out there either way.

Robert Grand is a Battalion Chief at Eugene Springfield Fire with 24 years of service. He writes Frontline Intelligence, a newsletter on operational doctrine, technology, and leadership in Fire & EMS.

If this landed, share it with someone in Fire & EMS who needs to hear it.