Predictive, Not Reactive: How AI Is Reshaping Fleet Safety and Maintenance

Fleet operators in both the public and private sectors are moving artificial intelligence from pilot programs into daily operations, targeting measurable gains in safety, uptime, and customer service.

At Guardian Bus Company, AI first entered the operation as a response to rising insurance costs and preventable incidents. Corey Muirhead, executive vice president, said the company deployed AI-enabled dash cameras to address driver behavior and reduce risk in real time.

“We’re using AI on two fronts: safety and customer service,” said Muirhead. “In relation to safety, we are using AI dash cams to monitor driver behavior. These AI dash cams were installed to reduce risk in real time and change driver behavior. Our insurance was skyrocketing, and our loss history was suffering from avoidable accidents.”

According to Muirhead, the impact was both immediate and long term.

“Driver behavior and performance have improved significantly. No headphones, no cell phone usage, less forward collisions and lane departure infractions,” he said. “Long term, we have seen a significant savings in insurance premiums. The company safety culture has completely changed.”

Guardian pairs AI dash cam data with GPS and telematics inputs to generate a driver safety scorecard. Recordable infractions are reviewed on a safety dashboard, with weekly and monthly reports analyzed by the safety team to identify trends. Drivers who are struggling are called in for retraining, and on-the-road instruction is used when needed.

“AI data is good, but the data is only as good as you use it,” Muirhead said. “The automation starts the process, but the human oversight still dominates the process.”

Beyond safety, Guardian is also using AI to manage call volume during peak school transportation periods. At the beginning of the school year and summer school, call centers are inundated with parent inquiries. The company now uses an AI phone agent that recognizes student profiles, cross-references route numbers with GPS data, and provides estimated arrival times.

“Parents deserve to know where their children are. If our AI phone system can help, that is our ROI,” Muirhead said, adding that the system is designed to supplement, not replace, live agents.

Looking ahead, Guardian sees route optimization and sick-call coverage as the most immediate opportunity for additional AI value. The company is building driver profiles so that when a driver calls out, the system can determine which available drivers can cover the route and push notifications with instructions.

“The simplification of this optimization will be a god send for transportation operators,” Muirhead said, noting the ongoing driver shortage and the complexity of daily dispatch decisions.

On the municipal side, the City of Long Beach, California, has expanded AI across its “1,600 rolling stock assets,” integrating Pitstop AI with Geotab telematics to shift from preventive to predictive maintenance.

“We wanted to move over from preventative maintenance to predictive maintenance,” said Eric Winterset, fleet services manager for the City of Long Beach.

After two years in pilot mode, the city deployed the platform fleetwide. Winterset said key performance indicators show unscheduled breakdowns are down 20%, towing costs have declined, and vehicle availability has increased to 93%, compared to a prior average of 90%.

To ensure AI insights drive action, the city has equipped supervisors and maintenance teams with customized dashboards showing the status of the vehicles under their oversight. Critical codes trigger email and text notifications, and technicians now review health reports during service visits to address non-critical issues before they escalate.

“This is just another tool in the toolbox for our staff and that is how we explained it,” Winterset said. “If it can help them be more productive, face less frustrating diagnostic situations, and working on a unit in a shop before it breaks down and know you are working on the side of the road were all appealing for them.”

Initial skepticism from staff was a hurdle. Winterset recalled an early conversation with a mechanic preparing for a road call on a no-start issue. The health report had predicted a low battery voltage and a potential no-start within three to five days.

“That prediction on that report was four days prior,” Winterset said. “He was stunned on how accurate the data was and then at the time, he became a believer of the product.”

For Long Beach, the next step is deeper integration between predictive diagnostics and parts management. Winterset envisions platforms that can identify the exact part number needed, check inventory, and initiate orders with contracted vendors before a vehicle is brought into the shop. He also expects AI-driven trend analysis to support proactive campaigns across similar units, such as replacing components across a model year before widespread failure occurs.

“I have been part of the AI journey into fleet for over 10 years, and it has grown so much but, I believe we are just scratching the surface of its capabilities,” Winterset said.

For both fleets, the message is consistent: AI is not replacing people, but reshaping how safety, maintenance, and dispatch decisions are made, with measurable operational results.