Three numbers for the same question: how much AI is there in companies?
The 88% quoted by executive surveys and the 20% measured by European official statistics do not contradict each other. They count different things, and knowing which to use changes the decision.
Executive summary
Executive surveys put AI adoption at close to 88% of organizations. Official statistical registers in the European Union put it at 20% of enterprises with ten or more employees. Neither figure is wrong.
The difference is not accuracy but universe: one asks executives who agree to answer about technology, the other surveys a representative sample of the business register. Contrasting them is what reveals where each market really stands.
The size cut is the most stable of all. In the OECD, 40% of firms with 250 or more employees used AI in 2024, against 11.9% of those with 10 to 49. The gap is more than threefold.
For a board the useful question is not how much AI exists in the market, but which universe your own organization is being compared against. Picking the wrong reference produces false urgency or false comfort.
Context
Since 2023 two families of AI adoption figures have coexisted. One comes from surveys of executives who agree to answer about technology; the other from the statistical surveys national institutes run on the business register.
The first describes those already in the conversation. The second describes the whole business fabric, with its workshops, its clinics and its distributors. When an investment plan leans on the wrong figure, the urgency diagnosis goes off.
Figure 1 · Self-reported AI adoption by region
% of surveyed organizations, 2025Table 1 · What each source measures
AI adoption in enterprises| Source | Universe | Year | Adoption |
|---|---|---|---|
| AI Index / McKinsey | 1,993 voluntary executive responses, 105 countries | 2025 | 88% |
| Eurostat, ICT survey | Sample of the EU business register, 10+ employees | 2025 | 20.0% |
| OECD, ICT usage database | Unweighted country average, 10+ employees | 2024 | 14% |
| OECD, firms with 250+ | Same universe, size cut | 2024 | 40% |
| OECD, firms with 10 to 49 | Same universe, size cut | 2024 | 11.9% |
Business implications
Set the universe before the target. Benchmarking against an executive survey means belonging to that universe: large organizations, with a technology budget and a willingness to answer.
Use the size cut, not the average. Adoption varies more between a thirty-person firm and a three-hundred-person one than between two countries on the same continent.
Ask for the methodology note before the headline. Field dates, response counts and the definition of "use" move the result more than geography does.
Methodology & data
Review of three public sources with declared methodology: Stanford HAI's AI Index 2026, which republishes McKinsey's 2025 global survey; Eurostat's EU ICT usage survey, with fieldwork in the first months of 2025; and the OECD ICT access and usage database, consulted in July 2025.
The AI Index and McKinsey figures come from the same survey: they must not be cited as mutual confirmation.
References
Stanford HAI (2026). The AI Index 2026 Annual Report, chapter 4. · Eurostat (2025). Use of artificial intelligence in enterprises, Statistics Explained. · Eurostat (11 December 2025). News release on the use of AI in EU enterprises. · OECD (2025). AI adoption by small and medium-sized enterprises. G7 discussion paper.