Why the Fire Service's Biggest Technology Leap Won't Come from a Tech Company
Robert Grand · Battalion Chief who still runs calls
Every major technology vendor is now selling AI to fire departments.
ESO is pitching it. ImageTrend is building it. Microsoft and Google are circling the space with enterprise deals. The demo videos look clean. The conference booths are impressive. The slide decks use words like intelligent, seamless, and workflow-optimized.
None of them have ridden in the back of a rig at 2am with a patient who is trying to die on them.
That’s not a small gap. That’s the whole problem.
The Documentation Burden Isn’t a Software Problem
For two decades, the fire and EMS industry has treated documentation as a software problem. If we just get the right fields, the right drop-downs, the right interface, crews will document better, faster, and more completely.
That thinking has given us progressively more complex PCR platforms that still get filled out in the parking lot after the call. Because the problem was never the software.
The problem is workflow translation, the gap between what happens in the field and what a structured data system is designed to capture. Tech companies optimize for clicks and completion rates. Field crews optimize for survival. Those two things are not aligned, and no amount of UI iteration closes that gap.
ESO has had years and hundreds of millions in investment to solve this. ImageTrend has built an entire ecosystem. The documentation burden in EMS has not meaningfully improved. Crews are still charting at 0300, exhausted, trying to reconstruct a call from memory into a form that was designed by someone who has never run a code.
That is a systems problem. And systems problems require someone who understands the system from the inside.
The Real Unlock Is Speech-Native Workflow
A few years ago, I was given a rare opportunity. As a Battalion Chief with 24 years in the fire service, I went back to work on the medic unit during a staffing shortage at Eugene Springfield Fire. I needed to see what the crews were actually dealing with. What I witnessed changed everything about how I think about this problem.
The paramedic techs were drowning. Not in calls, in charts. They were attempting to close out the last call’s documentation while simultaneously starting a new chart for the next one. On lower acuity calls where I rode in the back as the tech, I faced the same impossible choice every medic faces dozens of times a shift: talk to your patient, or document the call. You cannot do both.
After each call I’d sit down to chart and the shift was immediate. What started as I want to create a thorough, accurate record that reflects excellent patient care became how do I just get all the required boxes filled so billing doesn’t kick it back. And eventually degraded to the lowest possible standard: how do I close this chart without an error flag so I can move on.
That is what the documentation burden actually does to people. It doesn’t just waste time. It erodes the standard.
Then one day I saw one of the medics outside the ED, pacing around, apparently talking to his iPad. I walked over and asked him what he was doing.
“I’m using speech-to-text to build my SOAP narrative,” he said. “It’s faster.”
That was the moment. Not a product demo. Not a conference keynote. A medic walking circles in a parking lot because he figured out that talking is faster than typing.
150 words per minute versus 20.
That’s the entire gap. A medic who just ran a cardiac arrest can reconstruct that call in three minutes of talking. The same medic navigating drop-downs and free-text fields takes 25 minutes, and still misses clinical nuance because the form didn’t have the right field for what actually happened.
Speech-native AI documentation isn’t “talk to a chatbot.” It’s: say what happened, review what AI wrote, correct what it missed, sign it. That’s a five-minute workflow. That’s what post-call documentation should look like in 2025.
The reason it hasn’t been built well yet isn’t technical. The models exist. The speech-to-text accuracy is there. The reason it hasn’t been built is that you have to understand what a medic actually needs to say, the clinical shorthand, the protocol references, the situational context that never fits a drop-down, to build the architecture that turns a voice narrative into a compliant, billable, defensible PCR.
You learn that in the field. Not in a product roadmap meeting.
Every agency that signs a contract with a major EMS documentation vendor is doing two things: paying for a service, and generating training data for that vendor’s AI model.
Incident types. Protocol adherence patterns. Medication administration sequences. Outcome correlations. Response time data. The aggregate of how your agency operates, documented call by call, year by year, that is an extraordinarily valuable dataset.
ESO will use it to build ESO’s vision of a better system. ImageTrend will use it to build ImageTrend’s. That data doesn’t disappear, it just stops serving your organization and starts serving theirs. The AI that gets trained on your department’s calls will optimize for the average of every department in their network, not for the specific way your crews operate, the protocols your medical director wrote, the language your medics actually use in the field.
That’s the wrong direction entirely.
The right question isn’t who owns the data. It’s what the data is building. A system that learns your organization should make your organization smarter over time, not contribute to a generic model that serves ten thousand agencies equally and none of them exceptionally well.
What’s changed is this: the barrier to building the right solution has collapsed. AI-assisted development has made it possible for someone who truly understands the problem to code the fix. You no longer need a $50 million engineering team to build sophisticated software. You need deep knowledge of the problem and the drive to solve it.
That changes everything about who gets to build the tools that actually work.
The Insider Advantage
So my team and I built a prototype.
I’ve spent 24 years in the fire service. I’ve run the calls. I’ve written the after-action reports. I’ve sat in the battalion chief’s seat watching crews burn time on documentation when they should be resting, training, or going home to their families. I got frustrated enough that I stopped waiting for a vendor to solve it and started building the solution I’d always wished existed.
The goal was simple: get medics back to doing what they signed up to do, which is take care of patients, not chase billing fields through a form designed by someone who has never run a code.
What we built learns your organization. Not EMS in general. Not the average of every department in a vendor’s network. The same framework deployed in Eugene looks different than the same framework deployed in Texas, because the system shapes itself around how each organization actually operates. Your data stays with you. Every insight it generates belongs to your department. The AI gets smarter about your agency, not about someone else’s.
The technology that genuinely transforms fire and EMS operations won’t come from a vendor who learned about our workflows from a requirements document. It will come from people who understand that the unit isn’t called “the rig” in the drop-down menu, but that’s what every medic calls it. Who know that “working arrest” means something specific and consequential. Who have been in the room when an after-incident critique revealed a documentation gap that contributed to a bad outcome.
The fire service deserves technology built by people who’ve lived it.
That’s the bar. Hold every vendor to it.
Robert Grand is a Battalion Chief at Eugene Springfield Fire with 24 years in the fire service and the founder of First Responder Intelligence,”FRI”. His team is currently building Clear Run, a working prototype for AI-native EMS documentation, designed from the ground up by people who still run calls. If you want to follow the build or be part of shaping it, connect with Robert here on LinkedIn.
If this landed, share it with a chief who needs to hear it. The conversation in our industry needs to move faster.
Subscribe
From the floor, not the vendor booth. Two times a week.