AI in Supply Chain: Why Data and Context Matter

Artificial intelligence is dominating conversations in supply chain today. Every software vendor seems to have an AI story, and every executive is being asked about their AI strategy.

But amid all the excitement, many organizations are discovering that implementing AI is much harder than simply plugging a large language model into an ERP or supply chain application. Data is fragmented, business processes are complex, and without the right operational context, even the most advanced AI models can produce recommendations that are incomplete, impractical, or simply wrong.

That’s the main question discussed with Valentina Jordan, CEO and Founder of Nauta, during a recent episode of Talking Logistics.

With so much pressure on supply chain executives to develop a strategy and implement AI, I thought a good place to start would be for Valentina to discuss what companies are getting wrong as they rush forward.

Valentina reminds us that every project should involve people, process, and technology, but the pressure to implement AI is causing companies to dive directly into the technology and skip the people and processes behind it. She says companies often think of AI as “plug and play” without considering these other two factors.

“I think it needs to be seen more as a transformation than just as a technology that you’re plugging into organizations,”

The key point is that implementing AI isn’t simply another technology project. It can change how work gets done, which means companies need to consider the people and processes involved from the outset.

She describes the goal as creating a single source of truth, or an “operational brain.” This combines structured and unstructured data with the context, knowledge, and experience of operators so AI agents can use that information to make decisions.

Why does that matter?

“Decision-making is not just about the hard facts,” Valentina says. Without context, knowledge, and experience, AI is essentially being asked to provide a textbook answer rather than one grounded in how a particular company actually operates.

This is especially important in supply chain because it is fragmented by nature, with so many often-disconnected systems and stakeholders. A single shipment can involve suppliers, carriers, customs brokers, 3PLs, and others, while companies internally might be working across ERP, TMS, WMS, YMS, and other applications.

Valentina also addresses a common question: Should companies get their data foundation right before starting with AI? She says you’re never going to get it all right because of the chaotic and constantly changing nature of supply chains. Instead, she suggests starting with a specific problem where you can measure the results. Pick an objective, KPI, or process you want to improve, apply AI there, measure the impact, and evolve from there.

One question that often comes up is: Why not take ChatGPT or Claude and just plug it into your supply chain applications?

Valentina’s response is that the question isn’t necessarily whether companies can build AI solutions themselves, but whether they should. General-purpose AI tools have made it relatively easy to experiment and build applications. But moving something into production and maintaining it over time requires people to continually monitor it, provide feedback, improve it, and give it access to the right operational context.

As Valentina puts it, if you’re a manufacturer or distributor, do you really want every department maintaining its own ChatGPT or Claude connections?

There are governance considerations too, particularly around security and data privacy. Companies need to be thoughtful about what information employees are putting into public AI tools and how those tools are being used.

That doesn’t mean employees shouldn’t experiment. In fact, Valentina strongly encourages people to learn about AI, build things, and remain curious. As I suggested during our conversation, perhaps Claude or ChatGPT will become the equivalent of the Excel spreadsheets or macros employees have historically used for specialized needs alongside enterprise applications. Valentina agreed, while emphasizing the importance of using these tools responsibly.

Supply chain operations are inherently complex and multifaceted. How can AI address this?

Valentina comments that every company’s supply chain operations are different, with so many things arising as exceptions. That’s why AI needs more than general knowledge to operate effectively in a specific supply chain environment. It needs the context, knowledge, and experience of each operation to produce meaningful results.

She suggests thinking of AI like hiring a new employee. That person may arrive with skills and experience from a previous job, but they won’t be fully effective until they learn your company’s goals, processes, business rules, customer expectations, and ways of working.

The potential advantage of AI is that it can learn these internal factors as well as account for the many exceptions that occur in supply chain operations — something traditional deterministic workflows often struggle to handle. This is how it can help transform supply chain operations.

There’s another important dimension to this. In many companies, a great deal of operational knowledge isn’t documented at all. Experienced employees develop their own processes and ways of handling exceptions over many years, and much of that know-how resides only in their heads.

As Valentina explains, AI creates an opportunity to capture more of that operational know-how so it can become part of the organization’s institutional knowledge rather than residing primarily with individual operators.

That doesn’t eliminate the need for people. Instead, it can enable employees to spend less time on the work itself and more time applying their knowledge and judgment to make better decisions.

Source: Talking Logistics