# Germany Wants 80 Percent of Companies Using AI. Using It for What?

> Using AI is an activity, not a result. Count the problems it solved, not the people who logged in.

- URL: https://hsalem.com/using-ai-for-what/
- Author: Hassan Salem (https://hsalem.com/about/)
- Published: 2026-11-03
- Topics: AI, Germany, Management, Productivity

On Monday, Germany gave itself homework. By 2030, at least 80 percent of large companies should use AI.

My first reaction was not "too ambitious". It was a much smaller question: use it for what?

## Berlin put numbers on the table

The federal government's Digital-Gipfel opened in Berlin on November 2. The same day, the government's AI task force, led by the digital ministry, published its report "Deutschland als KI-Nation 2030" with 18 measures across ministries.

The 2030 targets are clear. At least 80 percent of large companies and 55 percent of small and medium-sized companies should use AI. AI-based output should be more than 10 percent of the economy. And 90 percent of federal agencies should use at least one AI application for core tasks.

The tone was urgent. The task force warns that the window to close the gap to the US and China "schließt sich bald", and calls the next 12 to 18 months decisive. Chancellor Merz called digital sovereignty in AI "a truly existential question" and said Germany must avoid becoming entirely dependent on third parties.

I'll be fair: I like that someone wrote numbers down. Germany already had a national AI strategy in 2018. Strategies are easy to praise and hard to check. Numbers can be checked.

## A login is not a result

But look closely at that one word, "use".

A company where three people in marketing share a chatbot licence uses AI. A company that rebuilt its whole claims process around a model also uses AI. In a survey, both tick the same box.

I have seen the small version of this in engineering teams. Leadership wants AI adoption, so someone builds a dashboard: licences assigned, weekly active users, share of pull requests written with AI help. The numbers go up. Everyone feels modern. A quarter later, nobody can say what actually got faster, cheaper or better for a customer.

That is not laziness. It is what happens when we measure the activity because the result is harder to see. There's an old name for it, Goodhart's law: once a number becomes the target, people start working on the number.

> No customer has ever paid for an AI adoption rate.

## Where this goes from here

My guess? "AI adoption" will show up in a lot of 2027 goals. Vendors will put the 80 percent on their first slide. And many Mittelstand companies will feel pressure to prove they "do AI", whether or not they found a problem worth solving with it.

There was one line from the summit I liked more than the targets. Digital Minister Karsten Wildberger said that as work changes, human judgment becomes more important. I agree. And judgment is exactly the part a usage target cannot count.

## Make your own number honest

You don't need to wait for 2030 to do this better in your own corner. Whether you write code, lead a team or are just deciding where your career goes next, here is what I would try this month:

1. **Swap the question.** Stop asking "do we use AI?" and ask "which problem did AI solve?" One sentence per use case: before, after, and how you know.
2. **Pick one boring, painful process.** Release notes, on-call handover, test data setup, the monthly report nobody wants to write. Time it today. Try AI on it. Time it again.
3. **Never put "AI usage" in a performance goal.** If you manage people, write the outcome instead. If you ask for usage, usage is what you will get.
4. **Keep a small log of where AI was wrong.** That is where your judgment earns its salary, and it makes a much better story in your next review than "I used it every day".
5. **Share the misses, not only the wins.** A team that only reports AI success stories learns very slowly.

A country can reach 80 percent by buying licences. A team cannot fake getting better. Pick the number that would still mean something if nobody was watching the dashboard.

Sources: [Bundesregierung: Digital-Gipfel 2026](https://www.bundesregierung.de/breg-de/aktuelles/digital-gipfel-2026-2457634), [ZDFheute: Merz fordert mehr Tempo bei Nutzung digitaler Technologien](https://www.zdfheute.de/politik/deutschland/ki-kuenstliche-intelligenz-bundesregierung-gipfel-100.html), [The Next Web (Reuters): Merz says Germany and Europe must build their own AI more quickly](https://thenextweb.com/news/merz-germany-digital-sovereignty-ai)