German companies use AI more than ever. Most still get little out of it

German companies have never used AI this much. And most of them get surprisingly little out of it. The ifo Institute found that 54.5 percent of German companies used AI in May 2026, up from 40.9 percent a year earlier. Yet 87 percent of AI-using firms with 20 or more employees say they capture little to none of the technology's full potential. The gap between using AI and gaining from AI is now the German AI story.

The numbers behind the gap

The Federal Statistical Office reports that 26 percent of companies with at least 10 employees used AI in 2025, up from 12 percent in 2023. Among large firms the figure is 57 percent, among small businesses 23 percent. The values differ widely by sector: from 12 percent in transportation and warehousing to 75 percent in information and communication.

The ifo Institute measures 54.5 percent in May 2026, against 40.9 percent in May 2025. Bitkom expects the German AI market to grow by 48 percent in 2026 to 28.7 billion euros, driven by autonomous agents.

At the same time, Mittelstand spending on digitalization fell by about 8.1 billion euros in 2024 to 23.8 billion euros. Firms finance their digitalization almost entirely from their own reserves rather than through credit.

People are already further ahead than many processes: according to the ARD/ZDF media study 2026, 37 percent of people aged 14 and over use AI tools such as ChatGPT at least once a week, up from 22 percent the year before. Among 14 to 29 year olds, 68 percent use them weekly. Only 13 percent of AI users have a paid subscription. And SISTRIX reports for October 2026 that Google shows an AI answer above the results for about 20 percent of searches in Germany.

Why tools alone do not raise productivity

Buying a chatbot subscription is not a workflow. Three patterns keep showing up: AI is used for single tasks, but not connected to the process. Nobody is responsible for the result. And nobody measures whether anything got faster or better.

Policy will not close the gap for you. The report Germany as AI Nation 2030 sets targets: by 2030, 80 percent of large companies and 55 percent of small and medium-sized enterprises should use AI. At the same time, the state spends about 800 million euros a year on AI projects, about 0.15 percent of the federal budget, according to the think tank Agora Digitale Transformation. Private AI investment in the United States stands at about 286 billion US dollars in 2025, according to the report. Do not wait for funding. Change your own workflow.

Where the payoff usually hides

First, one repeated workflow end to end. Quotes, invoices, customer replies or reports: pick one workflow that happens every week, and put AI where it really saves time.

Second, data in one place. AI can only work with what it can reach. If quotes live in one system, customer data in a second and appointments in a third, every AI answer stays patchy.

Third, one person responsible. Someone has to own the result and have time for it every week. Without a responsible person, AI stays a hobby on the side.

Fourth, staff who can check AI output. The ability to verify an AI answer is the real skill. A faulty quote that goes out unchecked costs more than the saved time is worth.

A four-week plan for a small team

Week one: pick one workflow and measure the time it takes today. How long does a quote, an invoice or a customer reply really take?

Week two: build the AI step, with one tool and fixed instructions. Not three tools. One.

Week three: run the new way alongside the old way and compare. What is faster, what is more error-prone, what is annoying?

Week four: keep what saves time, drop the rest, and write the new workflow down so everyone on the team does it the same way.

What your customers already do

Thirty-seven percent of people use AI every week, and for about one in five Google searches an AI answer already sits above the results. Your customers meet AI answers before they meet you. Make sure your website answers their questions clearly. How AI answers change your visibility is covered in our article on AI answers and your visibility. Why pilot projects often stall is covered in Why most AI pilots never reach daily work in Germany.

The takeaway

Using AI and gaining from AI are two different things. The numbers show it: use is here, the payoff is missing. The way there is not another tool. It is one workflow, one responsible person and one measurement. Start with one workflow, measure for four weeks, and then decide with numbers instead of feeling.