An AI pilot with defined evaluation criteria
Evaluate one AI use case with T-NEX: limited scope, suitable data and agreed criteria for expansion or stopping.
Explore serviceAI adoption in Germany
How many companies use AI? Understanding the figures and differences
Official statistics and industry association surveys cover different businesses, technologies and periods. The figures below therefore answer different questions. This comparison shows which value supports which statement and how to use it with its population intact.
In daily work
Which figure describes which businesses?
For businesses interpreting market adoption or using AI figures in decisions and presentations.
The overview distinguishes official statistics, a promotional bank’s research and industry association surveys. It groups the original sources assessed here by population. A report’s publication year is not necessarily the year measured by its adoption rate.
| Source and result | Population | Reference or survey period | Publication / data date |
|---|---|---|---|
| Destatis: 26% | Enterprises with at least 10 people employed in the ICT survey’s scope | Reference year 2025 | Table dated 24 November 2025 |
| KfW: 20% | SMEs, including micro-enterprises | Use during 2022 to 2024, KfW SME Panel 2025 | Focus 533, 11 February 2026 |
| Bitkom skilled trades: 4% | 504 trade businesses with at least 1 employee, weighted telephone survey | Weeks 23 to 29 of 2025 | Report published January 2026 |
| Bitkom AI: 36% | 604 companies with at least 20 employees, weighted telephone survey | Weeks 27 to 32 of 2025 | Report published February 2026 |
Overview
What the official figures show for Germany and the EU
Eurostat’s original table reports 25.97 percent for Germany in 2025, consistent with the rounded Destatis figure of 26 percent. The EU figure is 19.95 percent. The Eurostat survey covers enterprises with at least ten employees and self-employed people in the included economic activities.
These official publications are not independent surveys to average into an additional confirmation. Eurostat combines national surveys within the European statistical framework. The article used here states December 2025 as its data date and December 2026 as the planned update.
Overview
Company size changes the comparison group
The Destatis table reports 23 percent for enterprises with 10 to 49 people employed and 36 percent for those with 50 to 249. The largest reported employment group is at 57 percent. That final value is not the overall rate for the German economy.
A small trade business should therefore not use a survey beginning at 20 employees, or the rate for large enterprises, as its own industry figure. Choose comparison groups by size, activity and survey scope. A difference between groups does not by itself establish a cause.
Overview
A 2026 publication can measure earlier use
The KfW analysis published in February 2026 asks about use during 2022 to 2024. Focus 554, published on 29 July 2026, uses the same panel basis to examine technology types. The later publication is therefore not a new measurement of AI adoption in July 2026.
The two Bitkom surveys instead collected responses during 2025. A chart should not present these publications as a continuous annual time series. Assessing change over time requires a comparable survey series and attention to methodological changes.
Overview
Use, plans and impact are different measures
An adoption rate means an enterprise uses at least one covered AI technology. It does not automatically establish how many employees use it, how frequently it appears in workflows or whether business performance has improved. Planned use also does not belong in the share of current users.
Percentages within a group need their own denominator. A technology’s share among AI users must not be reported as a share of all companies. For open online surveys, also check who could participate and how potential self-selection was addressed.
| Statement | Information needed |
|---|---|
| AI is used | Which technologies and businesses were included? |
| AI is planned | Over which period, and in addition to which current users? |
| A use case has a high share | Of all respondents or only of AI users? |
| A productivity effect is reported | Measured change, self-assessment or modelled estimate? |
Overview
Quote figures without losing their meaning
A defensible short citation includes the publisher, population and reference year. For example: In Bitkom’s 2025 survey of 604 companies with at least 20 employees, 36 percent reported AI use. In a longer presentation, add the fieldwork period, method and a link to the original source.
Keep the sources separate. An average across skilled trades, SMEs and enterprises above a particular size threshold describes no known population. The range between these studies is not a margin of error for Germany either. Those differences are part of the finding.
Overview
Evaluate your own use case for the investment decision
An adoption rate can describe how widespread a topic is in a comparison group. It cannot determine whether document processing, knowledge search or a workflow extension will work in your business. That needs a limited task and a credible comparison with the current process.
T-NEX evaluates such questions through an AI pilot. Example data, quality criteria and handling effort are agreed beforehand. The outcome may support expansion, adjustment or another approach. These cited studies are external market research, not measurements of a T-NEX project’s impact.
FAQ
Frequently asked questions
Which official source is a useful starting point for Germany?
The Destatis table presents use by employment size. Keep its defined population and reference year. For a European comparison, add the harmonised Eurostat data.
Why is the skilled trades figure lower?
That survey covers a different set of activities and starts at one employee. Questions and survey designs also differ. The percentages alone do not establish the cause of the gap.
Do the KfW figures measure adoption in 2026?
The reports used here were published in 2026, but their adoption rate refers to 2022 to 2024 and the 2025 SME Panel. Publication year and reference period must be stated separately.
Can the studies be combined into an average adoption rate?
No. Their populations, methods and periods do not support that calculation. The resulting average would have no defensibly defined meaning.
Sources & further reading
- Destatis: AI use by employment size, 24 November 2025destatis.de
- Eurostat: AI in enterprises, data from December 2025ec.europa.eu
- Eurostat: Original 2025 tablesec.europa.eu
- KfW Focus 533, 11 February 2026kfw.de
- KfW Focus 554, 29 July 2026kfw.de
- Bitkom: Skilled trades digitalisation, January 2026bitkom.org
- Bitkom: Artificial intelligence, February 2026bitkom.org
Related options
You may also be interested in these.
AI automation for SMEs
Plan AI automation for an SME: select a suitable task, distinguish search from RAG, prepare data and permissions, and assess the complete workflow.
First version: Updated:
Read articleAI and software for skilled trades and SMEs
Handle enquiries, quotations and job documents more effectively. T-NEX develops office workflows and extends the software your business already uses.
Learn moreAI and software for industrial companies
Connect production workflows, technical documents and project portfolios. T-NEX works with operations and IT to define a practical first deployment.
Learn more