How AI Impressed (and Underwhelmed) at NAFA 2025

From the second I viewed the list of workshops scheduled at this year's NAFA Institute and Expo, it was clear that artificial intelligence (AI) would be the talk of the town. AI-centered seminars were so ubiquitous that, if someone really wanted to, they could learn more about the subject at almost every time slot dedicated to educational sessions across the event's three-day run.

I didn't take things to that extreme, but still attended a good number of AI-focused talks led by a variety of fleet technology companies. By the end of the conference, I was left with mixed feelings on how AI is currently being presented to fleets.

After some ruminating, here are my AI-related takeaways from NAFA 2025—the good and the not so good.

1. Certain AI fleet tools live up to the hype

Ever since the world got its hands on ChatGPT, every industry you can think of has been exploring the possibilities of large language models (LLMs). In the fleet space, technology companies have begun to incorporate AI-powered chatbots into existing solutions to give users more ways of interacting with them.

Until recently, it was a given that, if a fleet manager wanted to know something, their best bet was running a report. But now that AI chatbots have hit the scene, with some solutions, fleet managers can simply type "How much have I spent on repairs this year?" to have their question answered. And for folks who have never felt comfortable adjusting date ranges and toggling filters, the ability to gain insights about their operations without having to fiddle with reports could be a real game changer.

That being said, fleets would be well-advised to remember the limitations of LLMs before treating them as their go-to source for information. At a panel I attended on AI in fleet management, a representative from Geotab (who have their own LLM in Geotab Ace) noted how the adage of "garbage in, garbage out" is especially relevant to chatbots. If they're fed inaccurate or incomplete data, LLMs are known to output false or misleading information in a tone that conveys total confidence in what it's saying. These instances are referred to as "hallucinations" in the AI space and are the leading reason why fleet managers should probably treat fleet-aimed LLMs as assistants more than co-pilots for now.

Even with that drawback, it's worth emphasizing that fleet-aimed LLMs are already being used by fleets today. And given how much discussion of AI centers around theoretical applications, it's refreshing to see solutions delivering value to fleets in the real world.

2. Many applications of AI remain theoretical

Because AI effectively boils down to doing things with data, there's no shortage of applications that could theoretically take advantage of the technology. AI could optimize at-home EV charging by limiting electricity use to off hours, thereby reducing costs. By analyzing vast amounts of collision and driver behavioral data, AI could improve vehicle safety systems. You could even take things a step further by integrating AI into traffic signals and crosswalk signs to extend safety improvements to pedestrians as well.

I heard representatives from automakers and fleet technology companies riff on all of those ideas at NAFA 2025. And while they sound promising and might have great potential, at the end of the day, they're only concepts. In terms of being able to showcase how they're currently leveraging AI, the most tangible demonstrations these speakers could provide were their presentation slides.

Revolutionary technologies all have to start from somewhere, so I'm not suggesting that these novel applications of AI aren't worth exploring. I just think it's worth noting that, given the extreme hype around AI in and outside of the fleet space, that we're still in the nascent stages of putting AI to use.

3. Fleet managers may be wary of AI as a buzzword

The emergence of generative AI tools like ChatGPT has excited programmers and product managers more than anything in recent memory, and that enthusiasm has rubbed off on the solutions they develop. Nowadays, if you visit the website of any fleet technology company, odds are high that you'll find AI mentioned right on its home page. Many boast of "AI-powered" features that analyze vast data sets to identify trends and outliers humans would miss, with the implication being that such capabilities are only possible thanks to recent technological innovation.

But do these features actually hinge on AI or are their connections to AI being inflated to capitalize off a trend?

Fleet managers wanted an answer to that question at a workshop I attended on using AI solutions to reduce fuel waste. After hearing the presenter describe their solution's fuel efficiency monitoring capabilities as being AI-powered, audience members requested specifics on what AI technologies were underpinning the feature. After some hemming and hawing, the presenter conceded that the feature didn't employ anything like machine learning and instead only looked out for miles per gallon values beyond the norm (something entirely achievable without AI).

To be clear, the feature at issue sounded like something that could deliver real value to fleets. The only reason it received the pushback that it did was because of how it was mislabeled. The use of "AI" as a buzzword is what raised alarm bells.

Put simply, fleet managers are a perceptive bunch. They know the capabilities of the tools currently on the market, so they easily distinguish between marketing fluff and true innovation.

Fleet technology providers would be well-advised to remember that fact when invoking AI in their messaging. Without the ability to walk the walk and talk the talk, bandying about AI might be more of a detriment than a plus in today's environment.

---

At the end of the day, AI isn't going anywhere. Now that its current capabilities and future potential have been built up so much, there's no putting the genie back in the bottle.

What remains to be seen is how the fleet industry will integrate AI technologies moving forward.

---