ifo Study 2026: More Than Half of German Companies Now Use AI
For the first time, more than half of German companies use AI. An ifo analysis from 5 June 2026 shows 54.4 percent now use AI software, up from 40.9 percent a year earlier. For decision-makers, the question shifts from "Should we use AI?" to "How do we extract measurable value?". This article puts the figures in context, shows the differences by sector and size, and explains why high adoption is not yet value creation.
An ifo analysis from 5 June 2026 shows that 54.4 percent of German companies used AI in May 2026, up from 40.9 percent a year earlier. AI has moved from a topic of a few pioneers to a mainstream tool. ifo survey head Klaus Wohlrabe says AI has now definitively arrived across the breadth of the German economy. Industry leads with 58.7 percent, large enterprises stand at 67.2 percent, small firms at 51.2 percent and medium-sized firms at around 47 percent. AI is used most often for administration, data analysis, programming, written communication and information search. Yet high adoption is no proof of value: 33 percent of users report higher costs than expected, around 30 percent of GenAI projects are stopped after the pilot phase, and only 27 percent of companies train their staff even though 56 percent use generative AI. The tipping point therefore calls for a shift from experimenting to managing.
AI has reached the breadth of the German economy
More than half of German companies now use AI. An ifo analysis from 5 June 2026 reports 54.4 percent active use for May 2026, up from 40.9 percent a year earlier. AI has moved from a topic of a few pioneers to a mainstream tool, and the decisive question is no longer whether but how effectively you use it.
The basis is the monthly ifo business survey of several thousand companies, one of the broadest samples in Germany. A second independent survey, the Bitkom report from early 2026, shows the same trend of adoption doubling within a few years. How this creates pressure to act for smaller firms is covered in innobu's piece on the AI boom in the Mittelstand and the strategy gap .
Artificial intelligence has now definitively arrived across the breadth of the German economy. The momentum of adoption is high.
Klaus Wohlrabe, Head of Surveys at the ifo InstituteWhat the ifo numbers show
The ifo survey delivers more than a headline. Beyond active users, it measures how many companies plan or discuss adoption, making the further trajectory predictable. The pipeline is well filled and the share of holdouts is shrinking.
| AI adoption status | Share of companies | Meaning |
|---|---|---|
| Use AI actively | 54.4 % | majority reached |
| Plan adoption | 16 % | concrete intent |
| Discuss adoption | 21.6 % | under review |
| No adoption planned | around 8 % | shrinking minority |
The ifo figures are above the earlier Bitkom values from early 2026, because the survey took place later in the year and covers other sectors through the monthly business survey. Both sources paint the same picture: a fast, broad increase across all sectors.
Planners and reviewers combined, more than nine in ten companies have engaged with AI. Those still hesitating belong to a shrinking minority, and the gap to active users widens month by month.
Sectors and company size: where AI arrives
Adoption is no longer limited to tech firms and large corporations. Industry leads, but the remarkable part is the catch-up in sectors long seen as digitally reserved. The gap between large and small is also narrowing.
Notably, small companies are ahead of medium-sized ones, a sign of low barriers to entry with standard tools. In services, more than 40 percent use AI, and in construction every fourth firm now uses AI after previously marginal levels. This breadth shows that AI is no longer a sector phenomenon but a cross-cutting technology.
The Mittelstand is catching up: The real news is that small and medium-sized firms are closing in on industry and large enterprises. AI is becoming a base technology whose mastery decides competitiveness, not pioneer status.
What companies actually use AI for
AI is moving from experiments into everyday work. The most common uses are not spectacular but practical: administration, data analysis and communication. This proximity to routine tasks is exactly what explains the fast rise in adoption.
Administration and analysis
AI sorts, summarises and analyses large volumes of data. This is where value comes fastest, because the tasks are well defined and easy to measure.
Written communication
Drafts for emails, reports and texts are produced faster. Generative AI lowers the barrier to entry because it can be used without deep IT skills.
Programming and search
Code support and information search are further focal points. The entry through standard tools explains the speed better than large platform projects.
Wohlrabe points above all to advantages in routine tasks and the processing of large volumes of information. How this leads to real change rather than point efficiency is set out in innobu's piece on digital transformation with AI .
German and EU perspective
Rising use meets a tightening regulatory framework. From 2 August 2026, further obligations of the EU AI Act take effect, above all for high-risk systems. Anyone using AI broadly without building governance takes on a growing risk.
Breadth without governance is risky: With majority use, the number of applications that may fall under transparency and documentation obligations rises. Without documented responsibilities, the liability risk for management grows.
Two issues are especially pronounced in the European market. First, data sovereignty: many companies are reviewing European alternatives because they do not want to hand critical data to non-European providers. Second, data protection: any AI involving personal data must comply with the GDPR, from training data through prompts to outputs. Which deadlines and obligations are coming up is summarised in innobu's piece on the EU AI Act high-risk deadlines .
Challenges and risks: adoption is not value
High adoption figures are no proof of value creation. Several studies in 2026 show a widening gap between use and measurable return. The euphoria is giving way to a sober ROI review, and that is healthy.
The numbers mark three gaps. First, the cost gap: 33 percent of users report higher costs than expected, often through token consumption and compute. Second, the value gap: around 30 percent of GenAI projects are stopped after the pilot phase, the main reason being unclear business value. Third, the skills gap, which the TÜV training study 2026 quantifies.
Training lags behind use: 56 percent of companies use generative AI, but only 27 percent have trained their staff for it. On top of that comes shadow AI, tools used outside IT control without data protection. Both endanger quality and compliance.
A widely cited productivity paradox underlines the scepticism: in a controlled study participants felt around 30 percent faster but objectively took longer. Why measured ROI so often lags the perceived benefit is explored in innobu's piece on the AI productivity paradox and enterprise ROI .
What companies should do now
The tipping point calls for a change in strategy. Those already using AI should move from experimenting to managing. Concretely: measure, train, secure. Five steps help.
-
Measure ROI honestly
Set a clear baseline, define concrete use cases and include operating, governance and change costs. Only then does it become visible whether a tool truly delivers or just keeps people busy.
-
Invest in training
Train staff so they master the tools rather than just try them out. The gap between 56 percent use and 27 percent training is the biggest lever for more value.
-
Legalise shadow AI
Provide approved tools and clear rules instead of banning AI. This brings the use that is already happening into a secure, GDPR-compliant framework.
-
Prepare AI Act compliance
Identify high-risk applications, document them and clarify responsibilities before 2 August 2026. An early assessment prevents expensive rework.
-
Review data sovereignty
Decide where European or in-house models make sense for data protection and compliance and where established providers suffice. Keep models interchangeable to avoid lock-in.
Passing the halfway mark is a milestone, not a goal. The advantage now goes not to those who use AI but to those who turn use into measurable value: through honest ROI measurement, training and clean governance.
Further Reading
Frequently Asked Questions
According to an ifo analysis from 5 June 2026, 54.4 percent of German companies use AI as of May 2026, up from 40.9 percent a year earlier. For the first time, more than half of all companies use AI. A further 16 percent plan to adopt it and 21.6 percent are discussing it.
Industry leads with 58.7 percent, more than 40 percent of service firms use AI, and in construction now every fourth company uses AI after previously marginal levels. By size, large enterprises stand at 67.2 percent, small firms at 51.2 percent and medium-sized firms at around 47 percent. The gap between size classes is shrinking.
The most common uses are administration, data analysis, programming, written communication and information search. Generative AI lowers the barrier to entry because it can be used without deep IT skills. This proximity to routine tasks explains why adoption is rising so fast.
No. High adoption figures are not proof of value creation. 33 percent of AI users report higher costs than expected, and around 30 percent of GenAI projects are stopped after the pilot phase because the business value is unclear. In addition, only 27 percent of companies train their staff even though 56 percent use generative AI.
The tipping point calls for a shift from experimenting to managing. Concretely: measure ROI honestly with a clear baseline that includes operating and governance costs, invest in training, legalise shadow AI through approved tools, prepare for the EU AI Act obligations from August 2026, and review data sovereignty for critical workloads.