Around 40% of white-collar workers in the EU use AI on the job, against 6% in elementary occupations, according to a Joint Research Centre (JRC) brief that the European Commission promoted on Friday 23 October 2026. Its message for employers and trainers: AI upskilling is moving too slowly for small firms and mid-career staff.
Key facts
- The Commission’s Directorate-General for Employment, Social Affairs and Inclusion (DG EMPL) published “AI and skills in Europe – Closing the adoption gap” on 23 October 2026, summarising a JRC brief.
- About 30% of EU workers use AI tools in everyday work, and around 30% of enterprises with ten or more employees have adopted AI, the brief says.
- Up to 2025 the JRC found no clear link between AI exposure and unemployment, earnings or working hours.
- Short, flexible learning options for adults remain limited, especially for mid-career workers in highly exposed jobs.
- Finland’s National AI Upskilling Initiative has delivered more than 2,000 practical training sessions to SMEs.
Who is using AI at work, and who is not?
Exposure is highest in management, professional, technical and clerical roles. Most university graduates sit in the most exposed groups, while women are heavily represented just below the top band, mainly in clerical and secretarial work. Use is strongest in large, highly digital firms and in the Nordic countries.
The gap is the story. Many jobs that could use AI simply do not, and the JRC points to smaller companies and mid-career employees as the places where adoption stalls.
Two sources, two numbers
The JRC’s figure of roughly 30% of enterprises is higher than Eurostat’s latest official count. Eurostat’s survey of firms with ten or more employees put AI use at 19.95% in 2025, up 6.47 percentage points in a year. The size gap in Eurostat’s data is stark.
| Eurostat, 2025 | Share of enterprises using AI |
|---|---|
| All EU enterprises (10+ employees) | 19.95% |
| Large (250+ employees) | 55.03% |
| Medium (50–249) | 30.36% |
| Small (10–49) | 17% |
| Denmark / Finland / Sweden | 42.03% / 37.82% / 35.04% |
| Romania / Poland / Bulgaria | 5.21% / 8.36% / 8.55% |
The Commission page does not explain the difference. Reference years and definitions likely vary, and Eurostat’s next update is due in December 2026. Trainers quoting either figure should name the source.
Upskill or reskill: which jobs need what?
The brief draws a line that L&D teams can use directly. Highly exposed jobs that still depend on judgement, responsibility and human interaction, such as doctors, veterinarians and managers in professional services, are more likely to be enhanced by AI. Those workers need upskilling within their own occupation.
Other exposed jobs lean less on those human skills. General office clerks, keyboard operators and some finance roles face a higher risk that tasks are replaced, so the JRC says they need reskilling towards different occupations. That is a harder, longer route.
Age adds a second filter. A 2025 survey by the nonprofit Generation, run by YouGov across France, Ireland, Spain, the UK and the US, found that only 15% of workers aged 45 and over used AI at work, and that 38% of older workers trained each year compared with 50% of younger colleagues.
Where AI upskilling support already exists
The Commission points to the Pact for Skills and the Artificial Intelligence Skills Alliance (ARISA), which back sector partnerships, micro-credentials and pilot projects. Finland is the named example. Its national initiative targets the gap between large firms and smaller businesses with hands-on sessions rather than long courses.
The fit with Eurostat is telling. Finland already ranks second in the EU for enterprise AI use, which suggests that targeted SME training and adoption move together, although the brief does not claim a causal link.
What this means for trainers and HR teams
Three practical steps follow from the data.
- Map exposure by role, not by department. Sort jobs into “enhance” and “replace” groups using the JRC’s criteria, then plan in-role upskilling for the first and career-change pathways for the second.
- Design for the 45-plus learner. Short, practical modules tied to real tasks suit mid-career staff better than generic AI awareness courses.
- Use public channels. SMEs can join the Pact for Skills partnerships in their sector, and training providers can package courses as micro-credentials that employers recognise.
Earlier 3LNews coverage showed a similar gap between AI ambitions and training time in US firms, and the EU Skills Guarantee offers one model for reskilling workers whose roles are at risk.
FAQ
Does the JRC say AI is destroying jobs in Europe?
No. Up to 2025 it found no clear link between AI exposure and unemployment, pay or working hours.
Which workers need the most attention?
Mid-career staff in highly exposed jobs and employees of small firms, where adoption and training both lag.
Why do the enterprise figures differ?
The JRC cites about 30%, while Eurostat reports 19.95% for 2025; the sources likely use different years or definitions.
Sources:
- European Commission, DG EMPL – AI and skills in Europe: Closing the adoption gap (23 Oct 2026)
- EU Publications Office – JRC brief on AI and skills
- Eurostat – Use of artificial intelligence in enterprises (2025 data)
- Silicon Canals – Few mid-career and older workers use AI at work (Generation survey)
