Think Big, Start Small, Stay Human
Inside the real-world best practices shaping AI-powered fleet management.
Artificial Intelligence (AI) is everywhere! From voice assistants like Siri and Alexa, to online tools like ChatGTP, to traffic mapping solutions like Google Maps and beyond, AI is a big part of everyone’s day-to-day life.
For fleet professionals, the challenge isn’t whether to adopt AI anymore, it’s how to adopt it wisely. “The biggest mistake I see is people focusing on the technology instead of the outcome,” says Abhi Narayanan, Data Intelligence and AI Advocate at Geotab. “AI should not be the goal. Reducing collisions, lowering downtime, improving utilization—these are the goals.”
That sentiment is echoed by Dor Shay, CTO of Element Mobility and Autofleet, who warns against fleets bending their operations to fit rigid tools. “Any fleet executive choosing an AI solution should insist that AI adapts to their fleet’s needs, not the other way around,” he says. “Don't settle for a one-size-fits-all approach.”
Together, their perspectives paint a clear picture of what separates successful AI deployments from expensive science projects: outcome-driven thinking, clean data, transparent systems, and a firm commitment to keeping humans in the loop.
From dashboards to decisions
Fleet managers are no strangers to data overload. For years, the job has meant hours spent in spreadsheets, dashboards, and reports, most of it backward-looking.
“Historically, we’ve asked humans to do all the technical heavy lifting,” Narayanan says. “Sifting through massive amounts of historical data just to identify trends.”
AI changes that dynamic. Instead of forcing humans to hunt for patterns, AI models can process billions of data points, flag risks, and surface insights in near real time. But Narayanan is quick to stress that this doesn’t mean handing over decision-making authority.
“AI identifies the what,” he explains, “so that humans can better understand the why.” Shay agrees. In Autofleet’s route planning and vehicle assignment tools, AI handles the computationally intense work—optimizing routes, balancing demand, and assigning vehicles—but never without context.
“We surface potential risks directly in the UI,” Shay says. If a manager forces a new delivery into the schedule, “we show how that impacts the ETAs of other deliveries.” This visibility allows fleet managers to evaluate recommendations with full context, before deciding whether to implement a change or not. In other words, the best AI doesn’t replace judgment, it sharpens it.
Choosing the right AI
With so many “AI-powered” labels in the market, how should fleet professionals separate substance from smoke? Narayanan’s advice is straightforward: Start with the problem, not the product.
“You can’t just say, ‘We want AI for safety,’” he says. “You need to say, ‘We want to reduce collisions,’ and then evaluate whether the solution has the right data quality, scale, and maturity to support that.”
Data privacy, ownership, and security are equally critical, Narayanan adds. Fleet managers should be asking tough questions: Who owns the data? How is it anonymized? What certifications are in place to protect it?
Shay adds another dimension: flexibility. “Demand customization based on your actual workflows and constraints,” he says. “You should be able to see how the system makes decisions and adjust it as your needs evolve.”
This includes integration: An AI tool that can’t connect to your existing systems becomes just another silo. “If it doesn’t talk to anything,” Narayanan explains, “then it’s just another thing for you to manage.”
Man vs. machine
Both leaders are clear: The future of fleet management is not about autonomous decision-making. Rather, it’s about collaborative intelligence.
AI excels in two main areas, according to Shay. First, routine, high-frequency tasks, where automation can reliably monitor signals and trigger workflows without fatigue or error. Second, complex optimization problems, such as routing, forecasting, and vehicle assignment, where massive computation is required.
However,strategic decisions—those involving trade-offs, nuance, and organizational context—remain firmly in human territory.
Narayanan offers a safety example. If telematics and camera data detect signs of driver fatigue—lane drifting, yawning, erratic movement—AI can flag the risk. However, what happens next should not be an automatic, robotic, AI-driven response.
“A human can call the driver,” he says. “Maybe they’re 20 minutes from home, and you can talk them through that last stretch safely. That context matters.”
In this model, AI acts as an analytical co-pilot, predicting, alerting, and prioritizing, while humans decide when and how to act.
Data that actually matters
AI may be powerful, but it is not magical. Its effectiveness depends entirely on the data it’s given. As the old saying goes: garbage in / garbage out. “AI is only as good as the data it’s fed,” Narayanan says. For fleets, that data typically falls into three categories.
First is transactional data: fuel spend, maintenance invoices, toll payments, and other operational costs that link the vehicle and driver behaviour directly to your costs, he adds.
Second is telematics data. “This is your GPS, your fuel level, your engine health, and other vital elements from your vehicles as they operate day-to-day,” Narayanan explains.
Third is contextual data, which adds environmental and behavioural nuance: weather conditions, road risk, driver fatigue indicators, and other situational factors. During a recent snowstorm, Narayanan notes, Geotab data showed a 53% increase in collisions across affected regions.
Shay expands the definition further, emphasizing the importance of constraints. “The AI needs to understand not just what’s happening, but what’s possible and permissible within your operation,” he says.
Best practices
If there’s one theme both leaders return to repeatedly, it’s this: AI success is as much about culture as technology. Messy data is the number one enemy. Missing logs, disconnected systems or inconsistent definitions can undermine even the most advanced models.
“Data hygiene has to be elevated to the executive level,” Narayanan says. “Your data is a strategic business asset.”
Driver buy-in is just as critical. Without it, even well-intentioned AI initiatives can feel like surveillance. “Drivers’ first impression is often ‘Big Brother,’” Narayanan notes. “You have to frame this as a tool to get them home safe.”
Shay reinforces the importance of starting small. Don’t expect perfection on day one. Pick a single use case, measure ROI, refine, and expand. At Autofleet, that iterative approach helped Zipcar achieve a 71% reduction in downtime as models improved with richer data.
At Geotab, a focused effort on predicting battery failures led to a 95% success rate. “The result was massive savings on maintenance,” Narayanan says.
Finally, close the loop. AI systems must learn from real-world outcomes—whether a predicted repair was necessary or a flagged risk turned out to be a false alarm.
Think big, start small
AI is no longer optional in fleet management—but neither is skepticism. The fleets that win won’t be the ones chasing buzzwords. They’ll be the ones asking better questions, cleaning their data, breaking down silos, and insisting on transparency.
“Technology will keep evolving,” Narayanan says. “But humans will always be at the centre of this industry.”
Shay puts it more bluntly: “You can’t afford to ignore AI, but don’t get caught up in the hype either.”
The future of fleet management belongs to leaders who understand that AI isn’t here to take the wheel. It’s here to help us see the road ahead more clearly, and choose the best path forward.
