“Will our BOMs end up in ChatGPT?” Why AI anxiety in manufacturing is the real risk

One sentence that contains everything

“Will our bills of material end up in ChatGPT?”

I hear that sentence regularly. In meeting rooms of mid-sized manufacturers, from managing directors running a company in its third generation. From people who know every customer and can name every machine on the floor.

The concern behind it is legitimate. The outcome, which is doing nothing, is not.

While German manufacturers debate whether artificial intelligence endangers their know-how, competitors in Eastern Europe, in Asia and increasingly in their own market are using exactly that technology to quote faster, plan more precisely and produce more cheaply.

The uncomfortable version: AI is not the greatest risk to your competitive position. The decision not to use it is.

Why the fear makes sense

Let me be clear about where this comes from. I am not an evangelist telling anyone to push everything into the cloud tomorrow. I come from industry. I planned production lines, configured bills of material and implemented ERP systems before the word prompt had a technical meaning.

I understand why a managing director gets nervous when the sales team starts pasting customer data into an AI tool. Or when engineering uses a chatbot to draft technical text and nobody knows what happens to the input.

That nervousness is a sign of responsibility. It becomes a problem when it turns into paralysis.

Three fears, and what actually sits behind them

Three concerns come up again and again.

“Our data is our capital, and AI gives it away”

The most common one, and not unfounded. Anyone using a free consumer chatbot and pasting in design data, costings or supplier terms should know that this input can be used to train the model.

That is not an argument against AI. It is an argument for a usage policy.

Enterprise options exist today, from cloud services with contractual training exclusion to models running on your own hardware, where your data is never used for training. Your bills of material, your formulations, your costings stay in your environment.

The question is not whether AI is safe. The question is whether anyone has decided which AI environment is the right one here.

“Our IT specialist blocked everything, because of GDPR”

Also common, and again with a legitimate core. Data protection law sets clear requirements for processing personal data. Anyone deploying AI tools has to know where processing happens, who has access, and whether a processing agreement is in place.

But data protection law does not prohibit AI. It requires that you organise it responsibly.

A blanket ban on all AI tools is not a compliance strategy. It is a surrender. Because what happens in practice? People use the tools anyway, on private devices, without policy, without oversight. That is shadow IT, and that is the actual risk.

A structured approach looks different:

• Classification. Which data may go into which environment? Public, internal, confidential, strictly confidential.
• Tool approval. Which tools are cleared for which use case?
• Training. Do people know what they may enter and what they may not?

Note: I am not a lawyer and not a data protection authority. For the legal assessment of a specific situation, a specialised advisor is the right address. What I contribute is the operational bridge: how to deploy AI so that your data protection officer's requirements hold and the company still moves.

“We do not even have our master data under control”

My favourite objection. Not because it is wrong, it is often painfully accurate, but because it gets used as an excuse.

Yes, the material master is probably not clean. Yes, the supplier records contain duplicates. Yes, the bills of material carry inconsistencies nobody has touched since 2014.

And here is the thing: AI is good at exactly that problem.

Generative models are strong at recognising patterns in unstructured data, identifying duplicates and proposing corrections. The first AI use case in a manufacturing company does not have to be an autonomous warehouse. It can be cleaning up the material master.

Not glamorous. It is the lever that makes everything else possible.

What actually happens if you wait

Not a horror scenario. Three things I watch happen.

The competitor in your own market. A mid-sized supplier in southern Germany introduced AI-supported quotation costing. Response time on tenders went from five days to eight hours. The competitor in the same segment still needs five days. Guess who wins the business.

The skilled people who leave. Your best people, the production planners, the buyers, the design engineers, can see that other companies let them work with modern tools. For the next generation AI is not a spectre. It is an instrument they want to use. Ban it and you lose more than efficiency.

The slow erosion. No bang, no disruptive moment. Every quarter a little slower, a little more expensive, a little less competitive. Until the board asks why margin has been falling for three years.

What I recommend

Not a two-hundred-page strategy. A pragmatic sequence, built on nearly thirty years in industry.

Understand where you stand. A readiness assessment that starts with processes and data, not with technology. How mature is the master data? How digitalised are the core processes? Where are the quick wins?

Define the rules of play. A usage policy that enables people instead of blocking them. These tools yes, those no. This data yes, that never.

Start small and make it measurable. One concrete use case, ideally one that shows visible results within eight weeks. Not a revolution. A controlled pilot.

Scale, or stop. Not every use case will work. That is fine. The organisation will have learned how to evaluate, deliver and govern AI projects.

None of that requires a data scientist. It requires someone who knows how a material master record is built and understands what a language model can do with it.

Control instead of fear

Fear of AI in mid-sized industry is not a sign of backwardness. It is a sign that owners want to protect what they built.

But protection is not standstill. Protection means defining the rules yourself, instead of waiting for the market to dictate them.

Every quarter spent waiting is a quarter in which a competitor makes their processes faster, cheaper and better informed.

You do not have to do everything at once. You do have to start.

And the honest answer to the opening question? Your bills of material are probably in a chatbot already. Not because anyone decided it, but because a design engineer needed a quick answer last week. The question is not whether AI enters your company. It is already there. The question is whether you take control of it.

---

#AI #GenAI #MidSizedIndustry #Governance