Your AI, Same Question, Different Answers

Adding context is the game.

AI is no longer a side topic in our industry. According to Gartner, 75% of CEOs see AI as the technology that will most affect their industry. Yet in supply chain organisations, three in five use AI only for isolated tasks, and just 17% have managed to scale their AI pilots.

So most companies have tried it. Very few get real value from it. Why?

In September, at the SeaBlue Project Logistics Network and CrossTrades AGMs in Ho Chi Minh City, Thijs de Vos of AI Concepts Company and I ran two workshops called From AI Assistant to AI Colleague. Both filled within an hour: two full rooms of 35 forwarders each. We did not start with theory. We started with a question every project forwarder recognises.

How do you move a 75-metre wind turbine blade from a factory in Lille, via Antwerp, to a wind farm in Brazil?

We asked it twice. Same AI. Same question.

Round 1: the AI on its own

The answer was correct. Permits are needed in France and Belgium. Engage a specialist transport company. Find a suitable breakbulk vessel. Appoint a local agent in Brazil.

Nothing wrong with it. But nothing you can act on on Monday morning either. It reads like a textbook, because that is what it is.

And it can be worse. When we first tested a general AI on this kind of cargo, it suggested booking containers. For a 75-metre blade.

Round 2: the same AI, with context

Then we gave the AI the knowledge a seasoned project forwarder carries in their head and in their network. The answer changed completely:

• Permits: which authority in France and which in Flanders, the permit category for a blade of this size, and the warning to start the applications together with the quotation, because lead times run to several weeks.
• Road: the trailer class a blade of this length requires, the night-convoy window, and a fallback route permitted in parallel in case the first one is refused.
• Sea: what to check on the vessel, including crane capacity at the working radius and the port call at Sepetiba.
• Destination: which network partner in Brazil to call for a quote, and the customs points to watch before the cargo lands.

But the real difference came next. The AI did not stop at what it knew. It checked the answer against everything a complete, door-to-door solution requires, from the factory gate in Lille to the wind farm in Brazil, and flagged the gaps: which steps were covered, which were not yet, and where information or a partner was still missing.

That is what an experienced project forwarder does before sending a quote. It is also something an AI on its own simply cannot do. To see what is missing, you first need to know what complete looks like, and that only comes with the full context.

Same AI. Same question. An answer you can quote on, and a clear view of what still needs to be done before you do.

The lesson: the tool is not the advantage

Everyone has access to the same AI models. Your customers do, your competitors do, and so do the newcomers who want to take your business. The tool is fast becoming a commodity.

What is not a commodity is what you know. The lanes you have run for twenty years. The authority that always asks for one extra document. The partner in the port who picks up the phone at night. That knowledge is what turned a generic answer into a usable one.

This is also where the gap between leaders and followers is opening. Gartner reports that 80% of the supply chain leaders building AI into the core of their operations already see positive returns. They are not using a better AI. They are feeding it better context.

But what if your operations cover a whole global supply chain?

For a manufacturer, the core of the operation sits inside its own walls. For a forwarder, it doesn't. A single shipment like our blade touches a factory in France, a road authority in Flanders, a terminal in Antwerp, a vessel operator, a port and customs office in Brazil, and a wind farm at the end of the road.

No forwarder holds all that knowledge in-house. Your own documents and experience only cover your part of the chain. The rest sits with your partners: the agent who knows the permit officer at destination, the port specialist who knows which crane can really lift what, the colleague who did the same move last year. And that is usually where the gaps are.

So your AI can only be as good as the context your network gives it.

This is where we come in

Independent forwarders have always competed with the big integrators by working together. At Loxygen, as the knowledge centre of our networks, we help turn that cooperation into shared, structured knowledge: lanes, procedures, lessons learned and the partners behind them. That is the context a single company cannot build on its own, and it is what made Round 2 possible.

In the age of AI, a strong network is no longer only a source of business. It is the context that makes your AI worth using.

AI does not replace the experienced forwarder. It amplifies whatever you give it. Give it nothing, and you get the textbook. Give it the knowledge of your network, and it starts working like a colleague.

Three levels of context, and you can start at the first

1. Write it into your question. Who you are, the cargo, the Incoterm, origin and destination, what you need back. A two-line question gets a two-line-quality answer.
2. Give it your own documents. Your procedures, checklists, past routings and lessons learned. Most AI assistants let you attach these to a project, so you don't have to repeat yourself.
3. Connect it to your network's knowledge. This is the step no company can take alone. Through secure, open connection standards, your AI can look up trusted network knowledge only when you ask.

Gartner's advice points the same way: start with AI assistants before you invest in autonomous agents. You don't need an IT project to begin. If you can write a good briefing to a new colleague, you can write good context for an AI.

And your data?

That is the right question. More than half of supply chain organisations name data security, privacy and IP leakage as their main worries about AI. The answer is not to stay away. It is to stay in control: you decide what you share, with whom, and under which rules.

What to do on Monday

• Take one real task this week, such as an RFQ, a tender or a difficult customer email, and run it through the AI you already use.
• Ask it twice: once with a bare question, once with full context. Compare.
• Ask it a third question: "What is missing for a complete solution?" See what it can and cannot tell you.
• Name one person in your company who owns this, and pick one use case.

Join us live on 13 November

For everyone who could not be in Vietnam, Thijs and I are bringing the workshop online: 90 minutes for forwarding leaders and the people who will put it to work. You see the live comparison, try it yourself on one of your own documents, and learn in plain language what happens to your data. No IT background needed.

Friday 13 November 2026 · 09:00–10:30 CET · live online · 35 places, no recording

Register your place →

Geert De Wilde, Founder, Loxygen

Figures: Gartner, The AI-Native Supply Chain (2026), based on the 2025 Gartner CEO and Senior Business Executive Survey, the 2025 Gartner AI Adoption in Supply Chain Survey and the 2026 Gartner CSCO Role-Based Survey.