91��Ƶ

Transportation companies are using AI and machine learning tools to streamline maintenance and improve benchmarking. (Penske Transportation Solutions)

[Stay on top of transportation news: .]

Motor carriers, freight brokers and shippers are gaining access to a broad menu of emerging artificial intelligence capabilities from startups and established technology vendors, available both as bolt-on systems and embedded features in transportation management software.

This rapid proliferation of AI tools can leave transportation companies unsure of where to begin or what products and services are worth their time, and even tech providers themselves are constantly learning.

“AI is evolving at a pace that is faster than technology has typically evolved. Every six months, we’re having to relearn what AI’s capabilities are,” said Walter Mitchell, CEO of TMS supplier Tai Software.

Adding to the challenge is the fact that AI has become an all-inclusive buzzword and some vendors hype non-AI solutions as AI.

“Ranking loads by profitability or looking up a backhaul in a database doesn’t require AI. That doesn’t make the solution worthless, but fleets should be careful,” said Hans Galland, CEO of BeyondTrucks, another TMS vendor.

At the same time, carriers and brokers aren’t always well versed in machine learning and AI.

“They could look at something and think it’s AI when it’s just a really basic rules engine,” said Chadd Olesen, CEO of process automation firm AVRL. “AI should be self-learning. It should be pretrained. It should be implemented and able to learn by itself.”

Dylan Dameron, vice president of operations at third-party logistics firm Axle Logistics, breaks these latest technology advances into two primary buckets.

Image

Dameron

“There’s automation, and then there is AI,” he said, explaining that simple automation uses binary “if this happens, do this” logic, while AI can detect patterns, interpret text or generate responses. “There’s a million use cases between those two ends of the spectrum.”

Dameron said he recently heard from more than 100 AI and automation companies within one week, illustrating the influx of new technology capabilities in the transportation industry.

“We’ve been an early adopter with a lot of folks, but right now we’re doubling down on who we’ve had success with and leaning into folks that have a proven track record,” he said.

Identifying Problems to Solve

It’s easy to get distracted by the latest products and capabilities, but the most successful deployments begin with a clear pain point.

“You don’t go to the pharmacist and ask, ‘What’s your latest drug that I could try?’ You go to the doctor with your problem,” BeyondTrucks’ Galland said.

The more granular and specific users can be, the better, AVRL’s Olesen said.

“Sometimes people aren’t peeling the onion back enough to really go in and look at the root causes of the issues,” he added.

Image

Schrier

Doug Schrier, vice president of growth and special projects at McLeod Software, recommends carriers and brokers focus on an area with a high return and identify their objectives.

“Know what you’re trying to fix, [know] what the outcome should be and [know] how you’re going to measure that outcome,” he said.

Ben Wiesen, president of Carrier Logistics Inc., a TMS for less-than-truckload and final-mile operations, recommends beginning with narrow, achievable wins.

“Pick problems which can be solved today and utilize the expertise your vendors bring to the table,” he said.

Brian Work, president of AI communications firm CloneOps, said successful companies identify high-volume, low-complexity tasks, map them thoroughly, integrate AI into existing systems and set escalation rules.

Any process that is currently creating delays is a good candidate for AI implementation.

Eric Rempel, chief innovation officer for 3PL Redwood Logistics, said agentic AI is the next iteration of workflow automation.

“That’s what helps us manage operations, reduce our cost to serve and increase the bandwidth of what people can do,” he said.

Transportation technology provider Trimble has begun rolling out AI agents to eliminate manual bottlenecks in order processing, invoice scanning and breakdown response.

“It’s our job to properly diffuse these technologies … in the industry and make them appropriate for our customers,” said Jonah McIntire, Trimble’s chief product and technology officer.

Clearing Bottlenecks

Communication is a common choke point in logistics, and AI can help normalize unstructured interactions.

“Most delays come from fragmented phone calls, emails and portal updates,” said CloneOps’ Work. “AI turns conversations into structured data that flows directly into TMS, [customer relationship management] and insurance/compliance platforms.”

AI can read, classify and route messages. Tai’s Mitchell said it can detect new quote requests, create shipments, route messages to the right workflow and trigger follow-up automation to increase efficiency.

“We have one customer that was doing 40 carrier bills a day and went to 400 a day,” he said.

C.H. Robinson, the industry’s largest freight broker, receives more than 10,000 email price requests daily.

“People had to open each one, read it, get a quote from our Dynamic Pricing Engine and send the quote back to the customer,” said Mike Neill, C.H. Robinson’s chief technology officer. “Now an AI agent does all that autonomously. The customer gets an AI-derived price quote … in 32 seconds.”

Load tenders that once sat in queues for hours are processed in 90 seconds, even if the email includes 20 separate orders. Neill said the virtual agents can identify and retrieve any missing information.

“Our AI agents are unbelievable multitaskers,” he said.

McLeod Software has seen similar gains. AI reduced email-based order processing from five minutes or more to as little as 15 to 30 seconds. Some customers have reported saving 15 hours per representative per week through check-call automation, and manual track-and-trace operations have dropped 60%.

“Carrier reps’ productivity in terms of the number of loads they can book per week has gone up,” Schrier said. “Instead of the industry norm of 50 loads per week per rep, it’s gone up to 190.”

Bill Driegert, executive vice president of the Convoy Platform at DAT Freight & Analytics, said some brokers are using agents to negotiate prices or handle inbound calls.

Image

DAT's Convoy digital freight platform in use. (DAT Freight & Analytics)

The Convoy digital freight matching platform has integrated multiple AI tools throughout the platform and combines historical, real-time and contextual data to inform pricing and customer experience.

“We’re constantly on the back end, engineering it into the process,” Driegert said.

One reason AI is becoming so powerful is that it is no longer viewed as a single model, said Matt Cartwright, CEO of Magnus Technologies.

Image

Deep Dive

►Putting AI to Work in Trucking
►AI Adoption Raises Stakes for Data Security
►AI Rollouts Depend on People as Much as Technology

“Just like people have different strengths, you can stack multiple models — forecasting, optimization, classification, recommendation — and use them together to improve decisions across the entire order life cycle,” he said.

With more tools in play, orchestration is becoming essential. Project44 coordinates multiple AI agents — phone, email and task agents — so they work together to manage workflows.

“We manage the creation, optimization and tuning of agent prompts so customers receive the benefits of AI outcomes without being burdened by the technical complexity behind the scenes,” said Nick Ruggiero, director of product management at Project44.

Project44 most frequently deploys AI for data-quality improvements and exception handling, such as resolving inactive assets or missing equipment IDs and standardizing carrier-reported delay reasons.

Rethinking Exceptions

While many early use cases for AI focus on office workflows, visibility and exception management are maturing rapidly.

Chain deploys AI agents to handle check calls, review driver messages, log arrivals and departures, and flag exceptions early.

“Some clear opportunities lie in some of the most time-consuming tasks, like check calls, visibility tied to proactive exception management and email parsing/load building,” said Annalise Sandhu, CEO of AI-powered visibility firm Chain. “Everything else is noise until the low-hanging fruit is fixed.”

Image

El Khoury

Marc El Khoury, CEO of tech-enabled trucking company Aifleet, sees asset utilization and driver efficiency as even bigger opportunities.

“The average driver runs about 1,500 loaded miles per week, but they could be doing 2,000,” he said, citing data from the American Transportation Research Institute.

ABF Freight’s city route optimization, which uses prescriptive analytics to maximize equipment utilization and optimize routes, has resulted in more than $13 million in annual savings and significantly reduced planning time for the LTL carrier.

“What used to take the manager at our Baltimore service center four hours or more can now be done in just 45 minutes,” said Matt Godfrey, president of ABF Freight.

Magnus’ Cartwright said AI excels at “identifying cumulative opportunities,” such as smarter load sequencing, better return-trip planning, tighter forecasting and better handoff timing — that can add up when layered together.

“When AI can see the whole picture at once, it finds trade-offs and opportunities that are almost impossible for a human to calculate in real time,” he said.

CLI’s Wiesen noted that even automating data entry has secondary benefits. When information hits the system earlier, optimization engines have more time to work, improving both operational performance and shipper visibility.

Drawing on Data

AI relies on data and connectivity, so even the best models may fail when the underlying data is inconsistent, incomplete or siloed.

Image

Talwar

“There is no AI without good quality data,” said Rohit Talwar, senior vice president of software engineering for Penske Transportation Solutions.

Penske has deployed several AI and machine learning tools to streamline maintenance, reduce downtime and improve benchmarking, and has trained those models on more than 20 years of maintenance data spanning multiple equipment types. Examples include a prescriptive model that identifies patterns and helps technicians identify the best possible repair options and a proactive diagnostics model that combines telematics data, fault code data and operational data to predict when a truck could break down and advise customers on steps to solve the problem and reduce downtime.

Last year, Penske’s prescriptive model supported more than 200,000 repairs and its predictive model prevented over 95,000 road calls, resulting in cost savings, increased shop throughput and less downtime for its customers, Talwar said.

Meanwhile, Ryder System found early success by applying AI to two data-rich areas — safety and standard operating procedures. The company had years of structured data, such as incident timing and injury costs, and unstructured data, including accident write-ups, but much of it was difficult to use.

“Our first model was pulling that into an actual insights portal,” said Dave Yoder, group director of analytics and product innovation at Ryder.

Managers now receive daily, customized, actionable safety messages.

As more providers build AI tools, they increasingly rely on TMS data.

“If another AI system is coming along, it needs to access this information, and we have a responsibility to make it accessible,” Trimble’s McIntire said.

McLeod continues to build its own AI capabilities while supporting more than 150 third-party integrations. Embracing a dual strategy gives customers choices.

“Nobody can do everything, especially in transportation where there are so many niche applications,” said CEO Tom McLeod.

Redwood Logistics has addressed the same challenge with RedwoodConnect, its data unification layer that pulls from TMS, warehouse management systems and other sources to create a single flow across the platform.

But the same data that fuels AI can also hinder it. Legacy systems and years of accumulated data mean most companies need to clean and normalize data.

Still, many companies don’t have the volume of data required for robust machine learning.

“A singular company will need millions and millions and millions of datasets,” AVRL’s Olesen said, adding that companies also need the right type of data, such as the difference between cost and margin.

Looking Ahead

Companies not actively pursuing AI today should still be preparing to use it.

“Start looking at your data fields and the accuracy of your data,” said Joe Ohr, chief operating officer at the National Motor Freight Traffic Association.

Now is the ideal time to establish a data retention policy and clarify data ownership with vendors, he added.

McLeod recommends that companies jump in sooner rather than later.

“You better get going on AI because your competitors are using it,” he said. “Over time it will be difficult to compete without mastering these tools.”