AI Won’t Replace Your Officers. But It Can Bolster Your Back Office.
Jordan Wallach, Belfry, CALSAGA Network Partner
What five years of building software for security guard companies has taught us about where artificial intelligence actually helps, and where it doesn’t.
Every operator I talk to, whether they run 40 officers or 4,000, eventually gets to the same math problem. You win a new contract, and the field grows. Uniforms, training, scheduling, another site to keep covered. But you can’t grow the back office at the same rate; the margins in this business don’t allow for it. So the same three or four people who were managing 60 officers are now managing 90, then 120, often with no more hours in the day and no real increase in headcount. Eight, ten, sometimes more officers for every person handling scheduling, timekeeping, payroll, and billing behind them.
That imbalance isn’t new. It’s the defining constraint of this industry, and it’s the reason so many owners I meet are, frankly, exhausted by the parts of the job that have nothing to do with why they got into it. They didn’t start a security company to spend their evenings reconciling time sheets. They started it to protect people and property, and to build something their officers and their clients could count on.
For a long time, technology didn’t help much with that. Most of the software this industry inherited was never built for it. It was adapted from field service, adapted from general workforce management, stitched together from three or four disconnected tools that didn’t talk to each other. Scheduling in one place, payroll in another, a spreadsheet somewhere holding the whole thing together. I’ve sat in enough back offices to know that “the system” often meant three systems and a lot of manual re-entry in between. In California specifically, where licensing renewals, meal-and-rest-break compliance, and multi-agency reporting add real weight on top of everything else, that gap has always been more costly than it looks from the outside.
So when the industry started hearing “AI” everywhere over the last couple of years, I understood the skepticism. Most of what gets called AI in business software today is vague… a chatbot bolted onto a dashboard, a feature nobody asked for, a headline with very little underneath it. And in an industry built on accountability, where a client’s first question after any incident is “who was responsible for this, and can you prove it,” handing anything to a black box was never going to sit well. It shouldn’t.
The question worth asking isn’t whether a piece of software uses AI. It’s much narrower than that: does it take work off a specific person’s desk, does it operate under rules that person controls, and can they see exactly what it did and why. Framed that way, AI stops being a headline and starts being a tool. One that either earns its place in a back office or doesn’t.
We’ve spent the last year building toward that narrower bar with Belle, the AI teammate inside our own platform. One version of Belle watches shifts in real time and reaches out to officers on the exceptions that used to sit on a dispatcher’s desk all day: a missed clock-in, a post that’s gone quiet, a shift nobody’s confirmed. It operates under rules operators set themselves, and it escalates the moment something needs a human decision. It doesn’t get to decide what “resolved” means… the operator does. A second version answers plain-language questions about a firm’s own operational data, so a question that used to mean a report request now takes seconds.
The clearest proof point we have is a customer, not a claim. Jason Fellows built LincRo Security Solutions in Minneapolis past 120 officers without ever hiring a dedicated dispatcher. Belle has responded to 2,719 missed clock-ins on his account, and roughly 86 percent of those ended with the officer back on the clock without anyone picking up a phone. As Jason put it to us: “We can rest easy knowing Belle is watching the schedule for us. We let her do her thing, and if she escalates something to me or my team, that’s when we step in and start making calls. Until then, we know she has it covered.” That’s not a headcount reduction. That’s a firm that grew without needing to build a function it couldn’t afford, and a team that gets to spend its attention on the handful of things each day that actually need a person’s judgment.
I’d offer California operators evaluating any AI claim from any vendor, ours included, a short list of questions worth asking before you believe it:
Is it actually built on your industry’s data? Consider: shifts, posts, billable hours… or is it a generic layer dropped on top of software that wasn’t designed for how you operate?
Does it work under rules you set, with a clear, visible handoff when something needs a human, or is it a black box you’re asked to trust on faith?
Will the vendor show you real usage numbers from a real customer, not a projected outcome?
And does the pricing reflect the work it’s actually doing for you, rather than a flat tax on the size of your team?
None of this is really about AI. It’s about the same thing this industry has always needed from its technology: tools built specifically for how security firms operate, that give owners and operations managers back the hours they didn’t get into this business to spend on paperwork. That was true when we started Belfry five years ago, built entirely around this industry after months spent listening to owners, operations managers, supervisors, and officers about what was actually getting in their way. It’s still the mission behind everything we build now: automate the busywork, never the judgment calls that actually need a person.
Jordan Wallach is co-founder and CEO of Belfry Software, an all-in-one operating platform built for security guard services companies. He can be reached at [email protected].