By: William Jones
A former bank CEO, Mullins has built and sold three public-safety technology companies by spotting where a new technology’s costs and revenue would land before the market did. He thinks AI and real-time crime centers are next.
In the mid-2010s, as body-worn cameras spread through American policing, the argument was about the camera. Should officers wear them? When should they press record? Who gets to watch?
Kevin Mullins was doing math on hard drives.
Every officer on every shift was about to produce hours of video, all of it to be uploaded, labeled, stored for years under retention rules, redacted for records requests and kept under a court-proof chain of custody. The camera was a one-time purchase. The footage was a liability that grew every day it ran.
“Everybody was focused on the camera,” Mullins says. “The camera’s the cheap part. The footage is what you have to live with for years.”
The bill arrived quickly. In 2016, Clarksville, Ind., pulled its officers’ cameras after a new state law stretched retention from 30 days to 190. Police Chief Mark Palmer told the Associated Press that storage and maintenance had run $5,000 to $10,000 a year and would jump to $50,000 to $100,000 in year one. “That’s not the kind of thing we budgeted for,” he said. Neighboring Jeffersonville shelved its cameras too. By then, Mullins had already built a company around the problem.
It’s a pattern. Mullins spent roughly two decades in banking as a chief executive, chief financial officer and board director before moving into technology. He has since built and sold three public-safety technology companies and now runs SaferMobility, a safety and compliance platform the company says runs at more than 3,500 locations. Along the way he has made a habit of spotting where a new safety technology’s costs and revenue will land, often before the market does, and building the business that’s standing there when it arrives.
His edge is unglamorous. Most people who build for public safety come from policing or engineering, and judge a technology by what it does on day one. Mullins judges it the way a banker judges a loan: what it costs to carry, what it earns over its life, what happens when something goes wrong.
That lens has him looking at the next wave: AI, and the real-time crime centers pulling a city’s cameras, sensors and data into one room.
The Banker’s Arithmetic
Banking runs on records. Audit trails, risk assessments, compliance reviews and regulators who want proof rather than promises were the texture of Mullins’s first career. So was a particular arithmetic: what an asset costs to hold, what it earns, what it exposes you to.
He carried that into his first technology company, All Clear Systems, built on the idea that the bottleneck in police technology usually wasn’t the technology but adoption: helping agencies bring tools in and manage them. ACS was acquired by a private-equity-backed firm.
Proving the Video Is Real
His second company, QuadCrypt, followed the video. As more police work ended up on camera, hardly anyone was asking the obvious question: how do you prove a clip hasn’t been altered? QuadCrypt used blockchain to create a tamper-evident fingerprint for video, so evidentiary footage could be authenticated later and its custody chain traced.
QuadCrypt’s pitch came before “deepfake” was an everyday word. Authenticating video is now a live question for courtrooms, newsrooms and election officials.
“Once video becomes evidence, the first thing a good defense attorney asks is, how do I know this is real?” Mullins says. “I thought that question was going to get a lot harder, not easier.”
QuadCrypt was sold to private equity.
The Footage Is the Business
Intrensic was his answer to the hard-drive math. It sold body-worn cameras with the software, storage, workflows and evidence management agencies needed to handle video from upload to courtroom to records request.
The camera was the front door. The durable business was behind it: the recurring revenue from managing what it captures, year after year.
Intrensic was acquired by Wrap Technologies in August 2023, Mullins’s third exit.
Back to the Bank Branch
SaferMobility brings him back to where he started.
Under the Bank Protection Act of 1968, every federally insured bank must name a security officer and certify to its board once a year that the alarms are tested, the procedures written down and the program working. For years, Mullins was in that room.
Then a bank he helped lead was robbed. What stayed with him, he says, wasn’t the robbery but the attempt to reconstruct what happened. “We had a binder that said we’d done everything right,” Mullins says. “Then something real happened, and the question became: What can we actually show? That’s a very different question.”
SaferMobility is that question turned into software. Its platform lets banks and other organizations run safety and compliance procedures and prove they were done. SaferMobility shows that the routine preventive steps actually happened: a timestamped record of a particular branch on a particular Tuesday, not an annual signature saying the program works.
The Call at the Inflection Point
Mullins also advises companies across public safety, government, national security and the military, much of it through V2 Global. Prospective clients tend to call at the same moment: when they have promising technology but need a sharper strategy, a credible path into agencies or a more disciplined route to revenue.
In ballistic intelligence, he has focused on advancing on-site forensic analysis of ballistic evidence.
What clients are buying is range. Few have worked every layer of the business: building the product, financing it, pricing it, getting a police department to adopt it, and selling the company. Mullins has done every one of those jobs.
The Next Place the Money Lands
Real-time crime centers pull live feeds from cameras, license-plate readers, gunshot detection, 911 dispatch and drones into one room, with AI deciding what an analyst sees first. In 2020, the Electronic Frontier Foundation counted about 80 nationwide; by 2025, the National Real-Time Crime Center Association put the number above 300. When Goodyear, Ariz., police scoped one that year, the first phase alone was estimated at $1.4 million, a third of it software (West Valley View).
To Mullins, this is the body-camera moment again, at city scale. Agencies are buying for day one: the live map, the instant alert, the faster response. The business, and the hard questions, come later. Who owns the data? Who pays to keep it? And when a case goes to trial, can anyone show what the system did?
AI-written police reports are an early preview. Axon’s Draft One, launched in 2024, turns body-camera audio into a first draft of a report. The Palm Beach County Sheriff’s Office alone produced more than 3,000 reports with it between December 2024 and March 2025, and the Electronic Frontier Foundation found the tool did not retain the AI’s first draft. Prosecutors in King County, Wash., told police in 2024 they would not accept narratives written with AI help. California’s SB 524, in effect since Jan. 1, 2026, requires agencies to label AI-assisted reports, keep the first draft as long as the final one and maintain an audit trail.
“Agencies are going to buy AI the way they bought body cameras, for what it can do on day one,” Mullins says. “Then a case goes to trial, someone asks what the software did and why, and nobody can answer.”
Mullins says every agency or bank should ask an AI vendor three questions before signing: “Can I see exactly what the system did and what a human changed? Can I reproduce that a year from now? And who owns that record, me or the vendor?” The vendors with good answers to all three are likely to be the ones agencies can keep buying.
Civil-liberties groups argue records aren’t enough; the EFF says police shouldn’t use AI to write reports at all. Mullins doesn’t wave that away. “Auditability doesn’t make a bad system good,” he says. “But without it, you can’t even find out that it’s bad.”
The Signature and the Record
Somewhere this year, at every federally insured bank, someone will sign that paper again, attesting that the security program works. For most of the past half century, a signature was enough.
“A signature tells you somebody believed it worked,” Mullins says. “A record tells you whether it did.”
He’s betting the AI era will run on records. He has been early before.






