How AGS Pro Built AI Agents That Act on TrackTik 

How Pono Security Scaled a Multi-State Operation

A pioneering security firm layered reasoning AI on top of its system of record. The payoff: recovered revenue, sharper client conversations, and an operating model most competitors haven’t thought of yet. 

Most manned guarding companies are still circling AI from a safe distance. AGS Pro is not. The firm has built a working layer of AI agents on top of TrackTik, agents that don’t just flag problems but carry a decision through to a finished outcome, always with a person in the loop. 

For Lee Andrews, Founder & CEO at AGS Pro, AI began to augment the marketing side in 2023 and grew into something far more operational. The instinct was never to replace the team but to take pressure off schedulers and dispatchers who were already stretched thin. 

“The point was never to take anyone’s job,” Lee says. “It was to give our people back the hours they were spending chasing the same problems over and over.” 

Earning trust before it touched a schedule 

The first agent, built to recover coverage when an officer calls out or no-shows, ran in shadow mode for two weeks before it was allowed near a live schedule. The test was simple: would the agent make the same call a seasoned scheduler would? And at first, it didn’t. 

“The early runs only matched our team 30 to 50 percent of the time,” Andrews says. “I remember thinking, ‘ Am I wasting my time here? ‘ ” 

Rather than abandon it, the team rewrote the reasoning behind the agent, the logic it used to weigh skills, proximity, cost, and overtime risk. The next week, it cleared 90 percent. That 90 percent match against a real human decision became the bar every agent had to hit before going live. 

Today the coverage agent works a familiar rhythm. An officer misses a shift, the agent waits a set window, then reaches out by text. If the officer can’t make it, the agent surfaces qualified, available replacements for ops to dispatch, all without a scheduler having to notice the gap first. 

Agents, not features 

The discipline that makes this work is knowing when not to build an agent.  

“TrackTik already handles the deterministic work, the rules, the alerts, the open-shift posts,” Lee says. “There was no sense rebuilding that. At our scale it would have been far too expensive anyway.” 

So the team drew a line. TrackTik owns the deterministic workflows. Agents sit on top only where judgment adds something a fixed rule can’t, and only after iterating on which ones actually earn their place.  

“We’re not asking the AI to reason for itself,” Lee says. “We’re feeding it the reasoning our people already use, the human in the loop. Then it operates with approvals from the ops team.” 

That principle stretches well beyond scheduling. AGS is now building agents that generate client-specific training workflows in the LMS, draft post orders out of HubSpot, and distribute the right training materials to the officers assigned to each account. 

Why it all rests on TrackTik 

None of this works without a single, trustworthy source of truth, and that is the role TrackTik plays. Certifications, leave time, site requirements, billable hours, and incidents all already live in TrackTik. The fields exist, the workflows exist, and the agents simply read from and act on them. 

“The quality of the data in TrackTik is the whole reason this is possible,” Lee says. “It’s our data warehouse. Every agent reads from it. The data was a big part of why we chose TrackTik in the first place.” 

The contrast shows up across the rest of the stack. When the agents reach into QuickBooks, they’re working against a system that doesn’t yet have the same depth of structured operational data to draw on. 

What it caught 

The clearest proof came when AGS pointed an audit agent at a new contract. The agent compared the way the account was built in TrackTik, its positions and post orders, against the signed contract itself. 

It found a position where contracted billable overtime simply wasn’t being billed, across multiple weeks. In a separate check, the same agent caught an estimate that was about to go out underbilled by roughly $3,000, the result of an entire day of shifts that had been missed. 

“That’s money we would have just written off,” Lee says. “The agent caught it before it ever left the building.” 

Cold posts, non-billable overtime, and contractual compliance have become the pillars of that work: making sure the company gets billed accurately and on time for what it actually delivers. 

From there, AGS pushed the same idea toward the client relationship. A security council agent collects, reviews, and analyzes activity at each property, including local crime data, and turns it into something clients can act on in their quarterly business reviews. 

One example landed hard. Pulling three months of dispatch data for a single property, the agent found that 36 percent of overnight dispatches were going to one stairwell where people were smoking. The fix was a single camera. 

“The client could see exactly where their money was going, and exactly how to spend less of it,” Lee says. “Every client we show this to wants it. It speaks to their budget, their liability, their efficiency.” 

Where it goes next 

Lee sees the next phase as customization, agents shaped tightly around each company’s own workflows, and he’s clear about which vendors will be able to keep up. 

“The companies that win here will be the ones with well documented APIs who are willing to partner and actually build alongside their providers,” he says. 

For AGS Pro, that partnership with TrackTik is already underway. The agents are live and the savings are real.