Toronto-based transportation management system (TMS) provider Rose Rocket’s first foray into artificial intelligence (AI) didn’t go as planned. In 2017, before AI became entrenched in trucking industry lexicon, the company set out with a U.S. fleet to automate the entry of its bills of lading.
“We were trying to take a million bills of lading and using optical character recognition (OCR) to enter them into the TMS,” co-founder and CEO Justin Sky recalled in an interview with trucknews.com. “What we found was the entry clerk would always outperform the OCR model because of, very specifically, context.”
That order entry clerk had innate knowledge about the fleet’s operation and customers that the OCR couldn’t match. Fast-forward to today, however, with the rapid advance of AI technology and large language models (LLMs), an AI-enabled TMS can actually match that clerk’s contextual knowledge through machine learning.
“This whole generation of large language models is really quite special, and perhaps even underhyped,” Sky said.
Underhyped? Is that even possible? It seems everyone is talking about AI.
“I think AI 12 months ago was more buzzwordy than it is today,” said Bill Cain, director of product management at Trimble Transportation. “Today, it’s almost table stakes [for TMS providers] that you are bringing some level of an AI agent to help augment some of the business processes that customers do today.”
But, experts advise, don’t go using AI for the sake of using AI. First, advised Cain, identify a problem the technology can help resolve.
Identify a problem
“Don’t just use technology as a conduit to create a problem,” he advised. “Look in your business and ask, ‘What are the things that are keeping your leadership team up at night?’ Then we can work with you to help solve those business problems with technology.”
Maybe a fleet is struggling to reduce empty miles. Maybe its staff spends too much time on order entry. Maybe drivers are distracted by too many phone calls from dispatch while on a delivery. Maybe its accepting too many unprofitable loads, tying up capacity that could be better used elsewhere. These are the types of actual problems AI can help solve.
Hans Galland, CEO of TMS BeyondTrucks, looks at potential use cases for AI in three buckets: automation; decision making; and reducing risk and errors.
Many fleets, he warned, rush into AI adoption without a clear plan, something he refers to as “self-medicating.” He gave a real-world example of a food distributor with a US$65 million freight spend that wanted to optimize routing and navigation in hopes of saving $10 million.
“Our initial analysis, however, revealed a critical flaw: their drivers were given a week’s worth of routes daily and then freely resequenced stops,” Galland said. “This practice would have completely negated any algorithmic savings. We also discovered this ‘self-dispatching’ led to 100,000 customer stock-out events annually, costing them revenue and customer satisfaction.”
Simply using AI for route optimization would’ve been fruitless unless the underlying problem of drivers not following the given guidance was identified and resolved, he said.
Automation isn’t necessarily AI
Remember when electronic logging devices (ELDs) were first mandated, and there were suddenly hundreds of ELD vendors in the market, many of them raising buckets of cash from venture capitalists and promising to transform the industry? Where are they today?
A similar process is now underway in the AI space. At the very least, trucking companies need to understand the difference between native AI within a TMS, and a plug-in tool. Bolt-on products that typically identify and promise to resolve a single problem using AI aren’t necessarily bad, but it can become unwieldy to run a TMS with multiple plug-ins.
“An AI can only be so good as the data model that’s underlying it,” warned Sky. “If you make AI native, it’s just inherent. It’s part of the experience, not an afterthought.”
When assessing vendors, Trimble’s Cain suggested customers ask which LLM their product is built upon. “If they can’t answer that, it’s probably just more automation than anything,” he said. “If it’s making decisions versus just automating data through a workflow, then that’s probably more real AI than just fake AI. I believe AI is kind of invisible. You don’t really see it, it’s just there. It’s taking the data as it goes through, passing it through various agents that are going to perform algorithms that come up with an outcome that’s going to drive better business decisions for you.”
Trimble’s first use case within its TMWSuite TMS is its Order Module feature, which helps assess tender submissions to determine how well they fit into a fleet’s network.
“We built out an AI agent that allows you to do some tender evaluation as you’re trying to sift through hundreds of orders that are coming in,” Cain said.
The AI agent grades the freight based on its profitability and fit with the fleet’s available capacity so that fleet can very quickly decide whether to accept or reject the load, or farm it out to a third party.
“The algorithms we’ve built will say, ‘This is A Freight because your ability to execute is really high, but this is C Freight because the revenue margins are not really what you want. You set up some of the rules you’re looking for and as you accept that freight, it really helps you be more successful,” Cain said.
AI in the cab
Improved decision making and efficiency benefit everyone within a trucking company, not the least of which, the driver behind the wheel. An AI-enabled TMS can benefit drivers in several ways. It can reduce distracting phone calls from dispatch by more accurately guiding the driver to the right address, right down to the dock door.
And it can even steer drivers away from danger at 3 a.m. when the dispatcher and fleet manager are asleep and road hazards loom ahead. A soon-to-be-launched product from Trimble, for example, will alert drivers to weather events and suggest alternative routes.
“It will notify that driver very quickly based on the load and the route they’re on, give them a weather alert so the driver knows ‘I’m about to drive into a snowstorm.’ And then it will give them options to reroute around the weather or to reschedule their appointment,” Cain said, adding the tool will even predict the extent to which the driver’s speed will be affected by the weather event.
Trust the tech
Of course, that means everyone from the C-suite to the driver’s seat will have to put their trust into an AI’s decision-making ability. Rose Rocket’s Sky describes the three-stage process of implementing AI as: “In the loop, on the loop, out of the loop.”
The first phase involves actively participating and scrutinizing the AI’s decision making, followed by staying on the loop with continuous performance monitoring, before stepping out of the loop and entrusting the technology to make decisions.
“If you’re talking to a vendor and they’re saying your team’s out of the loop right away from Day Zero, that’s a warning sign,” Sky said.
BeyondTrucks’ Galland also urged fleets to be wary of AI tools that promise to be a quick fix, without understanding the root causes of the challenges the fleet is looking to resolve.
“I think what a lot of vendors today imply is that if you buy a bot or an agent, that it’s going to fix the problem,” he said. “No, it takes a broken process that was broken for a number of reasons and just automates that broken process. Essentially, you’re automating trouble into more trouble.”
Where is AI headed in the future? Sky foresees a day when fleets can effectively build their own customized TMS that grows and evolves with the business, no coding necessary. If a fleet begins handling more LTL freight, for example, their TMS will adapt to its new needs, eliminating the need to replace the TMS or run multiple systems.
The elephant in the room whenever AI is being implemented, is the potential for job losses. But Trimble’s Cain doesn’t see that as a real threat.
“There’s a lot of chatter out there that over the next three to five years all these roles are going to disappear,” he said. “I don’t think so. I think this is an opportunity to upskill a lot of our workers to learn new things, to be more technology driven than before — when they just came in, did their job and went home. Now, their job is going to be more driven around data analysis and cleansing of data to make smarter decisions, to make their lives easier, and that just opens up more opportunities for folks. I’d say, embrace the AI.”
