Input Stories

The “big unlock” provided by AI, says C.H. Robinson Senior Director for Commercial Connectivity Megan Orth, is its ability to handle unstructured data. Think of all the tools that supply chain professionals use to do business: email, text, phone calls, messaging apps, proprietary software and more. AI can digest all that information, learn from it and vastly improve the efficiency and accuracy of the supply chain process.

Big-time savings: C.H. Robinson’s AI agents have performed millions of shipping tasks—including delivering more than 1 million price quotes and processing more than 1 million orders.

Faster for small firms: Thousands of C.H. Robinson’s customers are small manufacturers, many of which use less-than-truckload shipping and place their orders via email.

• Whereas once these customers might have waited up to four hours for a human to process their orders, an AI agent now accomplishes those tasks in less than 90 seconds.
• “The AI reads an email, classifies an email, pulls out pertinent details and fetches any missing information from elsewhere in the system, then creates an order in the system,” said Orth, describing the process. 
• Another AI agent processes emails from carriers offering capacity on their trucks. Following its launch in January, it increased the number of trucks posted in C.H. Robinson’s system almost tenfold.

How it works: At C.H. Robinson, AI agents are built into the workflow with “humans in the loop.”

• The first step is uploading a lot of data, Orth said—providing access to all those emails, texts and so forth. Then, human operators teach the AI, guiding its decision-making by giving it lots of feedback.
• Over time, of course, that human intervention tapers off, and the AI agent becomes more capable and independent. 
• Yet, humans are never absent. The AI agents flag unusual cases so that a human employee can review them.

Freight classification: C.H. Robinson has built a fleet of more than 30 AI agents, Orth said, highlighting one in particular that determines freight classes  for LTL shippers. The National Motor Freight Classification system contains thousands of classes and codes, making it difficult for shippers to find the right ones. Often, the classifications they provide C.H. Robinson by email are either missing or incorrect.

• The AI agent needs minimal information like the correct dimensions and weight for the shipment to determine the correct classification—in about 10 seconds. A human would need 10 minutes or more to look it up.
• Even better, the AI agent can handle hundreds of such requests at once. All told, it’s saved more than 300 hours per day, getting more freight on the roads faster than ever.

Levels of AI: Orth distinguishes between several forms of AI, noting that manufacturers have been getting the benefit of traditional AI—like machine learning and advanced algorithms—for “more than a decade.”

• So what separates generative AI and agentic AI from previous tools? Traditional AI can provide services like predictive ETA, while generative AI can do things like answer emails.
• Agentic AI is another level up. These “agents” can interact with each other and make low-level, intuitive decisions—such as when the order-processing agent works with the classifying agent to get the correct order into the system.
• “Generative AI is sending the emails, but agentic AI is making the decisions about how to process the shipment,” Orth explained.

So how do you get started? For companies interested in building their own, custom AI agents, the process starts the same way: with an information dump.

• That information might come from dozens of different spots in the customer’s system, but as long as the AI has access, it can put everything together.
• This institutional knowledge is what enables the AI to be successful, Orth emphasized.

Set parameters: The next step is to find a business case and to set measurements for the AI’s success. Measurements are indispensable, Orth reinforced. “If you don’t have measurements, you will struggle” with implementing AI and getting value out of it, she said.
 
Onward and upward: C.H. Robinson is charging ahead in its adoption of AI, recently unveiling an even more advanced “ecosystem” that unites many of its previous innovations—and adds more benefits. The system “understands context, makes decisions in real time and self-optimizes global supply chains at scale,” according to the company.

• “For more than 120 years, this company has shaped the logistics industry. Today we’re doing it again,” said C.H. Robinson President and CEO Dave Bozeman. “Instead of chasing disruption, we’re engineering it and engineering it from within.”