The numbers, plainly.
Public figures on AI in small businesses. Every chart links its source.
The smaller the company, the less AI it uses
% of companies using AI
European Union, 2025
Eurostat, use of AI in enterprises (2025)Small firms, with no technical team of their own, fall behind.
| European Union, 2025, 10–49 employees | 17.0% |
|---|---|
| European Union, 2025, 50–249 | 30.4% |
| European Union, 2025, 250 or more | 55.0% |
| Spain, Q1 2025, Under 10 | 13.4% |
| Spain, Q1 2025, 10 or more | 21.1% |
Why they don’t use it
EU companies that considered AI and didn’t adopt it, 2025 (%)
- Lack of relevant expertise70.9%
How I handle it: I build it and train your team.
- Unclear legal consequences52.5%
How I handle it: AI Act notices and a GDPR agreement included.
- Data-protection concerns48.8%
How I handle it: Minimal data, in the EU, in your name.
An hour of work isn’t worth the same everywhere
Labour cost per hour, 2025 (€)
- Spain€26.4
- EU average€34.9
- Euro area€38.2
- Netherlands€47.9
- Denmark€51.7
A Dutch success story overstates the saving here. I calculate with Spanish costs.
Eurostat (31/03/2026), via Forbes EspañaWhat has actually been measured
Effect of generative AI in studies with real workers (% improvement). The metrics differ: compare sign and size.
Professional writing: less time per task Noy and Zhang, Science (2023)
+40Customer support, new staff: issues resolved per hour Brynjolfsson, Li and Raymond, QJE (2025)
+34Consultants, tasks within reach: speed Dell’Acqua et al., Harvard Business School (2023)
+25Customer support, everyone: issues resolved per hour Brynjolfsson, Li and Raymond, QJE (2025)
+14US workers who use it: work hours saved St. Louis Fed (2025)
+5.4Consultants, tasks beyond reach: correct answers (points) Dell’Acqua et al., Harvard Business School (2023)
−19
It helps a lot on some tasks and makes others worse. That’s why I test every automation on real cases.