By Ulugbek Ergashev, Chief AI Officer & Co-Founder, Datatruck

Last year, I watched three carriers with nearly identical fleets post margins that were more than fifteen points apart. Same lanes. Similar rates. Roughly the same headcount. The difference was not their drivers, their routes, or their equipment. It was their technology, and specifically how deeply AI was embedded in the actual workflows running their business.

I build AI systems for the trucking industry. I have seen this pattern repeat across hundreds of fleets, and it has become one of the clearest signals I track: the carriers pulling ahead are not necessarily the largest or the most sophisticated. They are the ones whose technology is doing the work that used to require a person at every step. And across the industry, AI tools are beginning to fundamentally change the economics of moving freight in ways that make that gap wider every month.

That shift is what this piece is about.

AI in TMS: Understanding What You Are Actually Buying

Walk the floor of any logistics conference and you will hear the word “AI” attached to nearly every TMS vendor’s pitch. That is not a bad thing. It reflects genuine momentum and real investment across the industry. But it does mean that carriers evaluating platforms need to ask sharper questions, because not all AI implementations are created equal.

There is a meaningful difference between a TMS that uses AI to automate genuinely complex, variable workflows like document reading, multi-board load matching, and real-time profitability scoring, and a TMS where AI plays a more supporting role in the overall workflow. Neither is dishonest, but they produce very different outcomes for the fleet operators running on them.

When you are evaluating a TMS, go beyond asking whether they have AI. Ask what the AI actually does inside the workflow. Ask whether it was designed around the real operational needs of carriers from the ground up, or built more broadly and adapted over time. The answers will tell you a great deal about what to expect once you are live on the platform.

What AI Should Actually Be Doing for Your Operation

The problems that AI must solve in trucking are not abstract. They are the same problems that slow down every dispatcher and every back-office team trying to move freight faster and at better margin. They start with documents.

Rate confirmations, bills of lading, proofs of delivery: the average carrier processes hundreds to thousands of these documents every week. Every manual keystroke is a potential error, and every error is a potential factoring rejection or delayed payment. AI that reads these documents accurately, every time, in seconds, does not just save time. It changes the economics of the back office entirely. That is not an efficiency gain. It is a structural cost reduction.

Load intelligence is the second place where AI earns its place. Finding the right freight at the right rate on the right lane is not a static problem. The market moves constantly, and manual load board searches are inherently limited to what one dispatcher can process in one sitting. At Datatruck, we built around this principle from day one: AI that simultaneously searches more than 100 boards, scores loads by profitability, validates counterparty risk, and books freight automatically does not just make dispatchers faster. It gives them access to a level of market intelligence they could never have assembled on their own.

Financial visibility is where I see the deepest pain. Most fleet owners know their top-line revenue, but very few can tell you, in real time on any given Tuesday, which lanes are systematically losing margin or whether the rate they are about to accept will be profitable after all costs are accounted for. That is not a data problem. It is a technology problem, and it has a direct solution in platforms that connect every operational data point as the business runs rather than at month-end.

AI-Native Is Not a Marketing Term. It Is an Architecture Decision.

I want anyone evaluating a TMS to understand one technical concept before they make a platform decision. There is a meaningful difference between a system built with AI at its core from day one and a system that has incorporated AI capabilities over time into an established architecture.

The difference is easiest to see in a single workflow. When a rate confirmation arrives on an AI-native platform, it triggers a chain: the document is read, the load is created, profitability is scored against current market rates, and the booking is confirmed, all without a dispatcher touching it. On a platform where AI was added later, the document might get read by an AI module, but a dispatcher still manually creates the load in a separate system, and the profitability scoring, if it exists, is a separate step that requires someone to run it. The AI is doing something, but the work still flows through a person at the handoffs.

That gap is not theoretical. It is where manual hours accumulate and where errors enter the operation. It determines whether your technology investment actually reduces headcount requirements or simply makes existing manual steps slightly faster.

Anyone evaluating a TMS today should come prepared with specific questions beyond the standard demo. Does the platform process documents natively and create loads automatically from rate confirmations, with no manual involvement? Does the AI search multiple load boards simultaneously, validate counterparty risk, and write the booking back to the TMS without a human touch point? Can you see profitability per load, per truck, and per lane in real time? Good vendors will welcome those questions. The answers will quickly reveal how deeply AI is embedded in the day-to-day workflow versus how much still depends on manual steps.

For carriers in particular, it is worth asking whether the platform was built primarily around fleet operations or designed to serve a broader range of stakeholders simultaneously. That is not a knock on multi-stakeholder platforms. Many do excellent work. But a platform where carrier workflows are the primary design consideration will feel different in practice, and the AI capabilities will reflect those priorities in ways that matter at scale. It is a conviction that investors are beginning to share as well. Our recent Series A reflects growing confidence that AI-native infrastructure built specifically for carriers is where the industry is heading.

The freight industry is in a period of meaningful technology advancement, and the companies that move toward AI-native platforms now are building operational advantages that compound over time. Faster booking cycles, lower administrative overhead, tighter financial controls, and the ability to scale without adding headcount. These are real, achievable outcomes, and more trucking companies are proving it every year.

The companies seeing the biggest gains are not necessarily the largest or the most tech-forward. They are the ones asking the right questions, choosing platforms designed for how they actually operate, and letting the technology handle the work that should not require a person in the loop.

Within the next 18 months, the gap between carriers running on AI-native platforms and those still managing freight through manual workflows will stop being a competitive edge and start being a structural cost problem. The economics are already diverging. Booking speed, load quality, back-office overhead, and margin visibility are all moving in opposite directions depending on the technology underneath the operation. The question carriers need to answer now is not whether AI belongs in their TMS. It is whether the AI in their current platform is actually doing the work, or just along for the ride.

—

Ulugbek Ergashev is the Chief AI Officer and Co-Founder of Datatruck, the AI-native transportation management system built for carriers. Datatruck serves more than 1,000 companies and has processed over $1.7 billion in freight. Learn more at datatruck.io.