Artificial intelligence (AI) in driver risk management is no longer a futuristic concept—it’s a present-day competitive differentiator. But with all the noise around AI, it can be hard to separate real, practical value from hype.
In our webinar, Cut the Noise: How AI is Actually Changing Driver Risk Management, Sean O’Bryan, Professional Safety Consultant at Verizon Connect, and Mark Beaumont, Executive Technical Consultant at The Hartford, brought the fleet operator and insurance perspectives to a candid conversation about where AI is delivering and where the industry still has work to do.
Here are the key takeaways.
SambaSafety’s 2026 Driver Risk Report, built from more than 50 million motor vehicle records (MVRs), 28 million telematics events, and a 13-year claims analysis, shows distracted driving is up 31% over the last two years. It also indicates there’s an 82.2% higher probability of a claim within 12 months if a driver exceeds the limit by 21-25 mph in a 60 mph speed zone.
Beaumont has seen these patterns accelerate at The Hartford, saying, “We see the drivers with their cell phones in their hands. They’re on speakerphone, and they think that’s hands-free, but hands-free is not distraction-free.” The data backs him up, indicating that phone use while driving increases crash likelihood by 240%, regardless of whether it’s handheld.
Driver fatigue, as O’Bryan points out, is a hazard that gets even less attention. Despite hours-of-service regulations, fleets have limited visibility into what happens before a driver’s shift begins. The Federal Motor Carrier Safety Administration (FMCSA) research indicates that fatigue is a factor in 13% of serious large truck crashes, and 65% of drivers say they occasionally drive while drowsy. “We have every opportunity as a fleet to do more when we think about training and educating our drivers on the risk of driver fatigue,” O’Bryan said. “It’s a silent hazard.”
The behavioral risks compounding losses aren’t hidden. They’re just not being caught early enough to act on. Traditional monitoring tools, such as annual MVR pulls and claims history reviews at renewal, were built around snapshots. But driver behavior doesn’t pause between check-ins. Violations build, patterns form, and by the time a risk control team sees the full picture, the window for early intervention has closed.
That’s the gap AI is positioned to close, and where O’Bryan and Beaumont say the real competitive separation is beginning to happen.
With 84% of insurance executives believing AI creates a competitive advantage and 83% of fleets saying AI is the future of safety, the insurance industry understands the value. But the most valuable applications aren’t the flashiest ones.
For Beaumont and The Hartford’s risk engineering team, AI’s biggest impact has been reclaiming time. “We’re using AI to get rid of the paperwork so we can spend more time with the customer,” he said. “The more face time we have with our customers, the better they perform.” With less prep and more presence, the technology clears the way for conversations that actually change outcomes.
O’Bryan points to prediction as AI’s most underappreciated capability. Verizon Connect’s analysis of a 34,000-vehicle case study found that in-cab alerts alone drove a 48% reduction in triggered events. And the patterns AI detects go beyond the obvious—something as routine as more than two rolling stops correlated with a two to three times increased likelihood of a collision. “AI is providing answers to questions fleet managers haven’t even thought to ask yet,” O’Bryan said.
That predictive capability of not only flagging what happened but also anticipating what’s coming is the shift forward. The difference between a reactive program and a proactive one isn’t the technology; it’s the timing. When AI can identify a pattern of behavior before it becomes an incident, risk control teams stop chasing losses and start preventing them. That’s the operational change organizations are beginning to realize, and it’s what separates the early adopters from everyone else still waiting to act.
Both O’Bryan and Beaumont are direct about where organizations get stuck, and it’s when they collect everything and act on almost nothing. “Too many fleets are placing an excess tax on their operations,” O’Bryan said. “They have every alert and every email turned on, and it quickly creates chaos. You ironically end up creating more exposure for yourself.”
Beaumont frames it as paralysis by analysis. “The systems will tell you who’s risky,” he said. “It’s what you do with that data that matters.” The technology is not the bottleneck; the business process is.
This is where the garbage-in, garbage-out principle becomes most consequential. AI surfaces patterns from the data it receives. Still, organizations that haven’t defined what “acting on a signal” looks like will find that AI amplifies their dysfunction rather than solving it. O’Bryan’s guidance is blunt. Less is more. The signals worth acting on share a few common traits:
That last point is where identifying a high-risk driver and changing their behavior diverge. O’Bryan is specific about timing: coach within seven days of an event. “The longer you wait, the more you normalize the risk,” he said.
The conversation itself matters just as much as the timing. O’Bryan’s framework keeps sessions to three to five minutes, starts and ends with something positive, and puts the driver at the center. “If your driver is not speaking for 60% of that conversation, you have it wrong,” he said. “The manager should just be there to provide the guardrail.”
Trust is the foundation on which all of it rests. O’Bryan recalled a logistics manager who invited drivers’ spouses to the technology orientation—not to sell the program, but to make the case why safety matters outside the cab. “If you want to build trust, extend your mission to the home,” he said. Beaumont described an executive team that installed both driver-facing and outward-facing cameras in their own vehicles alongside their drivers. “We’re asking this of you, and we’re doing it as well.” That level of visible commitment changes the cultural calculus around safety technology entirely.
As AI takes on more of the analytical heavy lifting, both speakers were unequivocal: human judgment cannot be automated away.
O’Bryan was direct about the legal stakes. Organizations that lean too heavily on automation without maintaining visible, documented engagement with their drivers are creating discovery risk they may not recognize until they’re in a courtroom. Reasonable supervision isn’t a cultural value; it’s a legal standard.
Beaumont put it in terms of accountability, saying, “It’s not about what you should have or could have known. It’s what you do know, and that’s where you need to take action and document it.”
That distinction matters more than most organizations realize. AI can process millions of data points, flag the highest-risk drivers, and recommend the right intervention. What it cannot do is sit across from a driver, understand the context behind a behavior, and build the kind of trust that changes how someone operates a vehicle. That’s a human job. The coaching conversation, the documented follow-through, the manager who shows up consistently—these are what transform a safety program from a compliance exercise into a culture. Organizations that blur that line by over-delegating to automation don’t just risk worse safety outcomes. They take on liability exposure that won’t show up until a claim is deep in litigation.
Looking ahead, O’Bryan sees edge computing reshaping how risk signals are processed with more power at the device level, faster signals, and less dependence on cloud-based evaluation. He also flagged privacy and biometrics as a legal frontier that will surface more frequently in litigation. “You have to be very, very careful about how you evaluate solutions when it comes to privacy and biometrics,” he said. “Understand where that information is being stored.”
Beaumont’s closing point was the simplest—and the one worth returning to: “AI is a tool. It helps us do our job better. It doesn’t replace the human element of how we can have an impact on improving fleet safety.”
The cost of ignoring that tool is measurable in claims that could have been prevented, coaching moments that went unaddressed, and behavioral patterns that quietly turned into crashes. The data, technology, and expertise to act on all of it are available to organizations that move fast enough to use them. As a result? They won’t just see better safety outcomes. They’ll see it in their loss ratios.
SambaSafety’s AI Profile Summary is now available to all platform users, giving managers a one-click, scannable view of every driver’s complete risk profile across MVRs, telematics, claims, and training records.