By Jason Truppi, Co-Founder and CTO of ForceMetrics

Read on Unite.Ai

Two years ago, if someone walked into a police chief's office and floated the phrase "artificial intelligence," the conversation would stall fast. Not because chiefs were closed to new technology (most agencies have spent the last decade digitizing everything from dispatch to records management), but because the early wave of AI tools promised more than they delivered. Vendors sold transformation. Agencies got tools that didn't hold up in the field, didn't fit real workflows, and didn't earn the trust required to stick.

That conversation looks different today. Chiefs aren't asking whether to explore AI. They're asking which problems it's actually good at solving, and how to deploy it in a way that holds up to scrutiny. That shift didn't happen because the hype got louder. It happened because a new generation of tools began solving real operational problems, and agencies got much sharper at distinguishing between a genuine capability and a demo.

What Actually Changed

The tools that are earning trust today share a common thread: they're not trying to replace judgment. They're trying to give it more room to operate.

The most useful AI applications in public safety right now handle the unglamorous work that consumes time without adding safety value: sorting through information, connecting scattered records, and summarizing what's already known so a person can act on it faster. These are assistive tools, not decision-makers. They reduce the cognitive load on people who are already stretched thin, and they leave the actual calls (who to charge, how to respond, what to prioritize) exactly where they belong: with the officer, the investigator, the analyst.

That distinction is what separates the tools gaining real traction from the ones that don't. Agencies have gotten better at asking the right questions during procurement: How is my data exposed? How does the system reach its output? Can we audit it? What happens when it's wrong? What models are best for what function? Vendors who can answer clearly are winning deals. The bar for what counts as "responsible AI" is higher than it was three years ago, and that's a good thing. It's forcing the technology to get better.

Where the Value Actually Shows Up

The honest case for AI right now in public safety isn't dramatic. It's about time, and about surfacing things that were always there but too hard to find.

Finding leads inside the noise. Investigators routinely sit on mountains of digital evidence, including body-worn camera footage, digital forensics extractions, case files, reports, and tip lines, far more than any team can manually review line by line. AI tools that can sort through that volume and surface what's actually relevant to a case are giving investigators hours back per case, and in some instances, turning up leads that a purely manual review would have missed simply due to volume.

Connecting the dots across reports -  Most agencies' records are scattered across disconnected systems, and the same person, vehicle, or location can appear in a dozen unrelated reports without anyone noticing the pattern. AI that can automatically group and connect people, places, vehicles, and incidents across those records, then surface the connection to an analyst or investigator, is turning fragmented paperwork into workable leads and helping close cases that would otherwise stall on disconnected information.

Enriching the record from dispatch through the courtroom -  One of the more underappreciated applications is real-time enrichment as a case moves through its full lifecycle, from the initial call at dispatch, through the response, through investigation, and into case management. When context is captured and enriched continuously instead of re-entered by hand at every handoff, the resulting record is more complete, more consistent, and better documented by the time it reaches a case file. That matters when that record eventually supports discovery obligations in court.

None of these is a speculative capability. Their workflows agencies are already running today, and the value shows up in the same currency every time: time returned to the people doing the work, and cases that move faster because the information didn't have to be found the hard way.

Doing More With Less

There's a structural reason this matters beyond convenience. The volume of digital and machine-generated data flowing through public safety agencies (video, sensor feeds, digital evidence, records from an ever-growing set of systems) will only keep growing, and it will keep outpacing agencies' ability to hire people to manually process it. Staffing has not kept pace with data volume in most agencies, and it's not going to.

That's the real argument for AI in this space. It's not about doing something flashy. It's about building the layer that can extract context and meaning from a volume of data that no team, no matter how well-staffed, could fully process by hand. Agencies that treat this as infrastructure, a system of intelligence sitting on top of their existing systems of record, will be able to do meaningfully more with the people and budgets they already have. Agencies that don't will find themselves accumulating more data every year with less and less ability to actually use it.

What Responsible Deployment Still Requires

None of this works if it's deployed carelessly. The agencies getting real value from AI share a few habits worth noting.

They start with the operational problem, not the technology. The question is never "how do we use AI," it's "what's actually slowing our people down, and is AI the right fix?" They insist on explainability: if a tool surfaces a connection or a recommendation, someone needs to be able to say why, in specific terms, not vague ones. And they keep the technology in an assistive role, surfacing information and reducing noise, not making the calls that carry accountability. Those calls stay with the people trained and empowered to make them.

The Conversation Worth Having

The most important shift of the last two years isn't that "AI" stopped being an uncomfortable word to say out loud in public safety. It's that the conversation got more specific. We've moved from "should we use this" to "which problems is it actually good at solving, and what does doing that responsibly look like?" That's a better conversation, and it's the one worth having going forward.

The agencies that get this right won't necessarily be the ones moving fastest. They'll be the ones whose deployments are still delivering and still trusted, five years from now. That's the bar worth clearing.

Author Bio: Jason Truppi is a technologist and entrepreneur with over two decades of experience at the intersection of public safety, cybersecurity and engineering. As a retired FBI Cyber Special Agent, he investigated some of the largest national security and criminal cyber intrusion cases, gaining deep insight into the challenges facing modern public safety agencies. Now as the founder of ForceMetrics, Jason leverages his expertise in machine learning, data analytics and software engineering to deliver scalable, data-driven solutions that empower public safety professionals to make better decisions and enhance community outcomes.