Feature, Freight News, IT Suppliers, Logistics, People

What Matters When Entrusting Supply Chain to AI

[ July 10, 2026   //   Gary Burrows ]

By Jordan Kass

When Transportation Management Systems first emerged in the 1990s, shippers were excited about the promise of technology that could operate their supply chains. But as with all tech tools, they quickly realized these systems weren’t exactly the easy button they hoped for. A TMS required a skilled team to configure it, run it and interpret and make use of its outputs.

Technology advanced over time. Visibility and analytics improved. Machine learning and algorithms helped automate some shipping tasks. But whether a shipper locked into a TMS or turned to a 3PL or 4PL, supply chains still relied on people coordinating complex global operations through handoffs, escalations, and disconnected workflows across the shipper and their various logistics providers.

That was the model the industry lived inside for 30 years. Now the newest forms of artificial intelligence have upended all the existing models, and the first technology that can autonomously operate a supply chain has finally arrived.

AI is available to everyone, but it’s not the same in everyone’s hands.

The latest breakthroughs make agentic AI capable of the reasoning and decision-making that supply chains depend on. But AI in and of itself is not all-knowing. While off-the-shelf AI applications can help you write a better email or chat with a customer or carrier, they don’t have all the data and deep expertise necessary to operate a physical supply chain – with all the implications for global commerce that implies.

Many people don’t realize AI requires access to a vast scale, breadth and richness of data to accurately make complex decisions. It’s not that more data is just better. It’s the difference between the AI functioning or not.

Lack of data is a prime reason why a lot of AI technology never gets beyond the pilot phase. It may have been a great idea, but the AI too often can’t decide. Or worse, it makes poor decisions or never reaches a level of accuracy you can trust.

Data + context is the essential formula for AI success.

Still, data alone is not enough. The context layer of AI is also crucial, and by context, I mean every minute detail you know about your customer and their needs from working with them for years, and the greater institutional knowledge that exists inside a 120-year-old logistics company like mine.

You can’t just build a generic AI system or even individual AI agents to manage shipments, because a generic shipment doesn’t exist. Every customer’s supply chain is unique, each of their products has specific shipping requirements, and their warehouses and loading docks have specific processes. Every shipping lane functions differently, depending on the season, the day or even the time of day. Each carrier you tender to operates differently.

The point is that AI needs to be deliberately and systematically trained on what the most talented logistics experts know, much of which isn’t conveniently written down somewhere.

When specific customer knowledge and deep domain expertise is encoded in AI, shippers get the benefit of infinite talent. The ability to serve them doesn’t depend on who’s awake in what time zone or when the person who knows the customer best gets back from vacation.

AI can provide high-quality, reliable, amazingly fast service 24/7 if logistics talent is built into it from the start. As the customer’s business changes and market and geopolitical conditions change, that also needs to continuously be fed into the system.

The first AI tech that operates, assesses and improves a supply chain.

At C.H. Robinson, our AI technology operates on more than 100 trillion proprietary datapoints derived from over 37 million shipments a year. That’s why it’s possible for 92 percent of our 4PL customers’ shipments across ocean, air, rail and trucking to be autonomously managed by our Lean AI Planner. We’re working on the edge cases to see how close to 100 percent we can get.

It begins at the order level, orchestrating mode selection, routing, consolidation, timing, tendering, tracking, exception management and settlement. Because the Lean AI Planner orchestrates across the full logistics lifecycle, it can consider tradeoffs among mode, service, cost, inventory positioning, carrier performance, and network efficiency simultaneously rather than in isolated workflows.

At the same time, our new Lean AI Engineer works in concert with the Lean AI Planner to continually assess a shipper’s supply chain and improve its performance. Together, they are one connected system that uniquely enhances a supply chain as it runs.

The Engineer analyzes the outcomes. It identifies where cost could be saved, where service is degrading, where carrier behavior shifted, where network instability is emerging, and where orchestration logic should adapt. Every insight generated by the Engineer improves orchestration inside the Planner. The next order benefits from what the previous order taught it.

It’s an exciting time for logistics and an exciting time to explore just how far AI can take us.

Jordan Kass is president of managed solutions of C.H. Robinson.

Tags: C.H. Robinson