AI Provides a Predictive Edge for Fleet Maintenance

A Knight Transportation mechanic working on the undercarriage of a truck. (Knight Transportation)

Today, artificial intelligence is driving tangible improvements in Knight-Swift Transportation’s fleet and maintenance operations, including predicting parts failures and preventing as many as 15 tows a month. Just three years ago, however, AI really was not a focal point for the industry’s largest truckload carrier.

In that span, Knight-Swift and other trucking companies have embraced AI as a core operational tool and in some cases are already realizing a measurable return on investment.

“We have over 100 AI projects that we’re looking at and experimenting to find out if there’s a return,” said Steve Grover, senior solutions engineer at Knight-Swift.

Mark Kennedy, the fleet’s corporate vice president of equipment, said the push to utilize AI in the trucking industry has been rapid and hard to miss.

The tone at major industry events also has changed noticeably, said Kennedy, who is general chairman and treasurer of American Trucking Associations’ Technology & Maintenance Council.

At TMC’s 2026 Annual Meeting in Nashville, Tenn., in March, “everybody was talking about it,” he said, noting that it’s a contrast from just a few years earlier when AI rarely surfaced in discussions.

Phoenix-based Knight-Swift ranks No. 7 on the Transport Topics Top 100 list of the largest for-hire carriers in North America.

As businesses implement various forms of AI across the supply chain, fleet maintenance has emerged as one of the earliest proving grounds for rapidly evolving technology.

TMC plans to devote additional attention to AI at its Fall Meeting in Pittsburgh in September, including establishing task forces to develop recommended practices for the technology’s use. One potential step involves applying coding for vehicle maintenance reporting standards (VMRS) to AI-driven insights, TMC Executive Director Robert Braswell said.

Image

Braswell

“I think there is a sense, if you don’t jump on the bandwagon, you’re going to be on the table as lunch,” he said.

That sense of urgency is driven by a growing belief that early adopters are building a competitive edge through data.

Equipment Health

Fleets already using AI tools are gaining a clearer view of equipment health, said Rocco Marrari, vice president of sales for Pedigree Technologies. The company’s Predictive­View platform, for example, can flag potential failures two to three weeks in advance.

Marrari estimates the industry is 18 to 36 months from AI-driven maintenance recommendations becoming standard, particularly among midmarket fleets.

The shift underway extends beyond identifying problems to recommending actions.

Russ Daniels, senior director of marketing at Cox Fleet, said AI can incorporate maintenance history, operating conditions, parts availability, shop capacity and cost considerations to guide next steps.

For fleets, the long-standing challenge has not been collecting data but making sense of it.

AI is beginning to close that gap by delivering more consistent, actionable decision-making across locations and operations.

‘Whisper Before the Scream’

Some technology providers say AI can detect early signs of failure before traditional diagnostic systems do.

That includes identifying degradation signals before a fault code appears, effectively “catching the whisper before the scream,” said Ganes Kesari, founder and CEO of predictive maintenance software vendor Tensor Planet.

Generative AI translates technical outputs into plain-language recommendations technicians can act on, helping improve adoption at the shop level, Kesari added.

“Instead of cryptic fault codes, technicians get, ‘Schedule a forced regen on Unit 247 within a week to clear the early soot accumulation detected,’” he said.

That clarity matters, Kesari said, because insights only produce ­value if they lead to action.

In one deployment with a Maine-based waste hauler, Tensor Planet’s platform reduced exhaust-related repairs by 41% within 10 weeks, saving about $1,600 per truck, he said.

Those results reflect a broader shift, Kesari said.

“Fleets today are making real maintenance decisions based on AI predictions,” he said. “The ROI isn’t theoretical anymore. The challenge now is adoption speed, and as an industry we are making progress.”

Beyond Maintenance

Fleets are also expanding AI into other areas of operation.

At Knight-Swift, drivers experiencing roadside breakdowns first interact with an AI system that gathers details and logs the issue before transferring the call to a human agent, reducing hold times.

The carrier also uses AI to prequalify driver applicants. Previously, staff members handled more than 1,500 weekly leads. Now, an AI chatbot contacts applicants within minutes, asks qualifying questions and routes only viable candidates to recruiters.

Other providers are applying AI to performance benchmarking.

Penske Transportation Solutions’ Catalyst AI platform analyzes large volumes of fleet data and compares performance against similar operations across its customer base.

Samantha Thompson, vice president of customer success and fleet telematics, said 67% of customers using the system improved key metrics such as fuel efficiency, utilization and maintenance adherence within the first month.

In one pilot, a fleet operating more than 6,500 power units identified underperformance relative to peers and reduced idling, resulting in six-figure weekly fuel savings.

Penske Logistics ranks No. 12 on the for-hire TT100.

Even as use cases expand, fleet executives are approaching AI implementation deliberately.

The ROI isn’t theoretical anymore. The challenge now is adoption speed, and as an industry we are making progress.

Ganes Kesari, Tensor Planet

Image

Pedigree’s Marrari said fleets new to AI should focus first on identifying specific pain points, defining success criteria and scaling gradually as results are proven.

A Distributed Approach

Knight-Swift, for example, has taken a distributed approach by testing AI tools across different parts of the organization rather than deploying a single system companywide.

“There was a drinking-out-of-a-fire-hose type of scenario,” Ken­nedy said. “We had to make sure we were addressing the most important things first.”

That approach also can help limit risk if a system fails or underperforms.

Still, industry leaders say the next challenge will be determining how many AI applications or platforms fleets truly need.

With a growing number of vendors offering subscription-based tools, executives must decide where the technology delivers the most value.

“Do I need 10 of these solutions, or will two do?” TMC’s Braswell said. “Maybe that’s a big part of the learning curve.”