Europe’s AI gap: people use AI. Most companies still don’t.

People are ahead of their employers

Generative AI reached European workplaces through the front door and the back door at the same time. In 2025, 32.7% of people in the EU aged 16 to 74 used a generative AI tool in the previous three months, according to Eurostat. Among working adults aged 25 to 64, the figure was 36.1%, and 21.9% used these tools for work.

Companies are moving more slowly. In the same year, 20.0% of EU enterprises with ten or more employees used at least one AI technology. Only 5.4% used AI to automate workflows or support decision-making, the kind of use that changes how an operation runs.

The two surveys measure different things, so the numbers are not a like-for-like comparison. People were asked about generative AI tools; companies about any of eight AI technologies. The direction, though, is clear: much of the AI already inside European companies arrived with employees, not through a company decision.

That matters because AI used person by person is AI nobody manages. A global study by KPMG and the University of Melbourne found that 70% of employees who use AI at work rely on free, public tools, while only 42% use tools provided by their employer. When that happens, the company controls neither the data going in, the cost of running it, nor the quality of what comes out.

Adoption is accelerating, but size decides who benefits

The pace is real. The share of EU companies using AI rose from 8.1% in 2023 to 13.5% in 2024 and 20.0% in 2025, more than doubling in two years.

But company size still decides who gets there first. In 2025, 55.0% of large companies (250+ employees) used AI, against 30.4% of medium-sized companies (50–249) and 17.0% of small ones (10–49). Large companies have data teams, IT budgets and room to experiment. Mid-sized companies, the backbone of European commerce, usually have none of these, even though their processes are just as repetitive and their margins often thinner.

1. Denmark42.0%
2. Finland37.8%
3. Sweden35.0%
4. Belgium34.5%
5. Luxembourg33.6%
6. Netherlands33.2%
7. Austria29.9%
8. Germany26.0%
9. Estonia23.4%
10. Slovenia21.6%
11. Malta21.5%
12. Lithuania21.3%
13. Spain20.3%
14. EU average19.9%
15. Ireland19.6%
16. France18.2%
17. Slovakia18.0%
18. Czechia17.6%
19. Italy16.4%
20. Croatia15.2%
21. Latvia12.2%
22. Portugal11.5%
23. Hungary10.4%
24. Cyprus9.3%
25. Greece8.9%
26. Bulgaria8.6%
27. Poland8.4%
28. Romania5.2%

Source: Eurostat, isoc_eb_ai. Enterprises with 10+ employees, excluding the financial sector.

The gap between countries is even wider. In Denmark, 42.0% of companies use AI; in Finland, 37.8%; in Sweden, 35.0%. At the other end, the figure is 8.4% in Poland, 8.6% in Bulgaria and 5.2% in Romania, the lowest in the EU. A Danish company is eight times more likely to use AI than a Romanian one. For companies in the slower half of Europe, this is both a warning and an opening: competitors abroad are already building the advantage, while local rivals have not started.

Commerce: growing fast, but not yet in operations

Commerce is moving faster than the average. Among EU wholesalers, AI use jumped from 14.1% in 2024 to 23.4% in 2025. Retail rose from 11.3% to 15.5%, and transport and storage from 8.1% to 11.2%.

1. All sectors13.5%19.9%
2. Wholesale trade14.1%23.4%
3. Retail trade11.3%15.5%
4. Transport and storage8.1%11.2%

20242025

Source: Eurostat, isoc_eb_ain2. EU27, enterprises with 10+ employees.

Look at what the technology is used for, and the picture changes. Only 5.6% of EU wholesalers use AI to automate workflows or support decisions, and just 1.8% use it for logistics. In retail, 2.4% use AI for logistics; in transport and storage, 2.1%. Almost twice as many wholesalers use AI for marketing and sales (10.1%) as for automating their workflows.

In other words, most AI in commerce today writes text. Very little of it takes orders, chases suppliers, plans deliveries or answers customers inside the systems where the work happens. That is where the time and the margin sit.

Why companies hold back: it’s not doubt, it’s implementation

Eurostat also asked companies that had considered AI, but do not use it, what stopped them. Only 17.8% said AI is not useful for their business. The main obstacles are about making it work:

1. Lack of relevant expertise70.3%
2. Unclear legal consequences53.6%
3. Data protection and privacy concerns52.7%
4. Availability or quality of data43.5%
5. Incompatible with existing systems41.6%
6. Costs seem too high38.4%
7. Ethical considerations24.6%
8. AI is not useful for the company17.8%

Source: Eurostat, isoc_eb_ai. More than one reason could be given.

70.3% lack the relevant expertise. More than half are unsure about the legal consequences (53.6%) or worried about data protection (52.7%). Around four in ten point to data quality (43.5%), incompatible systems (41.6%) or costs (38.4%).

None of these is a technology problem. The models exist and are widely available. What is missing is the work of fitting them into a real company: knowing which processes to start with, connecting them to the systems already in place, handling data responsibly and knowing what it all costs.

What this means for commerce leaders

1. Map before you buy. Start with a clear picture of how your teams work, which AI tools are already in use and which workflows take the most time. Most companies have never done this, and it is the fastest way to find where AI pays off.
2. Bring individual AI use into the open. Your people are probably already using AI. Replace scattered personal tools with shared, approved ones, so you control the data and the cost.
3. Start in operations, not only in marketing. Order entry, supplier follow-up, customer queries and delivery coordination are repetitive, measurable and close to the margin.
4. Measure from day one. Track what each automated task costs, including AI usage, and what it saves. Without measurement, there is no way to know what works or what to improve.

Europe’s AI gap will not close by buying more tools. It will close when companies turn the AI their people already use into systems that are implemented, connected and measured.