What is ahead for manufacturing companies?

In the last months I have met several people from other companies to discuss the changes AI brought to their business. All agreed that it is a game changer, but there is a big difference in how it is approached.

Software companies (especially the bigger ones) are becoming AI-first very quickly. For them it is already part of everyday work. They are rebuilding processes and teams and redefining positions. Anthropic recently published a report about how AI changed work inside their own company, and their AI agents already work six times more hours than their people.

Why manufacturing is slower

Companies producing physical goods are much slower. Especially the small and middle-sized ones.

Everyone there knows that AI is a big thing. But they are really busy doing their business. There are orders to deliver, customers to serve and problems in production to solve. There is simply no time or people to spare for experimenting with AI.

Also, using AI tools comes more naturally to a software developer than to a mechanical engineer, production technician or quality engineer. A developer works with text and code every day, so a chatbot or a coding agent fits into their work quite easily. For an engineer who works with drawings, machines and materials, the connection is not so obvious.

And there is one more thing. I don't see any clear signal that would force them to change. If their sales drop, it does not come with a label saying "you should have used AI". It may look like a bad year, a stronger competitor or pressure on prices. They may continue doing business as usual, not knowing what went wrong.

Will they become AI-first companies?

The most obvious answer is that they will transform into AI-first companies as well. I am not so sure about it.

From my experience, changing the culture and processes of a company is very, very hard. And it is not something you can simply buy. Hiring one AI specialist will not do it. Such a person usually finds out quickly that the tools are the easier part of the job, and that changing how people in the company work and think is much harder.

And then there is a question I don't know the answer to. Do they really need to become AI companies at all? Is there still a place for a company that does not rely on AI? Maybe such a company would keep more freedom and independence. But it may also lose its customers to competitors who are faster and cheaper. Then the freedom does not help much.

Will AI-native startups take their business?

In software, it is now much easier to build a new product with a few people. A small team with AI tools can do work that needed a whole department a few years ago. So software companies have a real reason to watch new competitors.

In manufacturing the situation is different. You still need capital for machines and buildings. You need certifications, which take time and money. You need a supply chain and, most importantly, customers who trust you. Customers usually don't order critical parts from a company that was founded last year. That is not something you build quickly, and cheap code does not help much here.

So I think this moat protects manufacturing companies quite well. At least for now.

Will the tools come to them from outside?

The third option is that they will be transformed by tools brought in by third-party integrators. Maybe they don't need to become AI companies at all. Maybe they just need to learn to use tools that make their work easier.

There is a historical parallel I find interesting. When electricity came to factories at the end of the 19th century, the factories did not become electric companies. Somebody else built the power stations, and the factories bought motors and machines. Most of them never had to understand electricity in depth. It came to them inside the products they bought.

I think AI can reach many manufacturing companies in a similar way. Through their ERP system, their CAD software, the machines they buy, the next version of the tools they already use.

But with electricity it was not so simple. The economist Paul David described it in his paper The Dynamo and the Computer (1990). The first electrified factories gained almost nothing. They replaced the big steam engine with a big electric motor and kept everything else the same: the shafts, the belts, the layout of the whole building. The real gains came only in the 1920s, about four decades after the first power station, when factories were redesigned around a small motor on each machine. New layout, new flow of material, new way of working.

If this applies to AI as well, buying the tools will not be enough. Somebody will still need to rethink how the work is done. And this is maybe where integrators can help the most: helping to change the process around the tools, not only installing them.

The moat is not infinite

I guess we will see sooner rather than later which way it goes. Probably it will be some mix of all three.

AI progress still shows no signs of slowing down, and it is just a matter of time before it steps out into the physical world and starts to transform this kind of business as well. The moat of these companies is big, but it is not infinite.