AI and the future of VET: who gets the chance to learn?

Predictions about artificial intelligence and work often swing between two extremes: AI will take everyone’s job, or it will make us all more productive. The reality is likely to be more complicated and different across occupations and sectors.
A new report from the European Training Foundation, Changing landscape of skills in the age of AI, offers a welcome starting point for thinking about what this means for vocational education and training (VET). Its central message is that AI is changing the tasks people perform and the skills they need. But how those changes play out depends on decisions made by employers, educators, workers and policymakers.
Changing tasks, not just jobs
AI can automate some tasks, help people perform others and create new work altogether. In a Cedefop survey across 11 EU countries, 67% of workers using AI said it helped them do some tasks faster, while 41% said they were doing some new or different tasks. These are reports from AI users, not evidence that jobs as a whole have disappeared.
The changes will vary. Administrative workers may use AI to process information or draft documents. Manufacturing technicians may encounter it through predictive maintenance. Healthcare workers may use systems that support decisions or record-keeping. And while practical work in unpredictable environments still requires human skill, digital tools may change how that work is organised.
For VET, this means looking closely at what is happening in each occupation, rather than assuming there is a single “AI future of work”.
The problem of getting started
One especially important question is what happens to young people entering work. Many entry-level jobs involve relatively routine tasks. These may not be glamorous, but they give beginners a chance to practise, learn from colleagues and gradually take on responsibility.
If AI takes over some of those tasks, where will that experience come from? And how will newcomers learn to check AI-generated work if they have had few opportunities to do the work themselves?
The evidence on employment is mixed. A Banque de France analysis found a sharper decline in permanent hiring of young people in occupations highly exposed to generative AI. But it could not establish that AI caused the difference. Meanwhile, Eurofound’s EU-wide analysis found no evidence of an AI-driven employment shock specific to young workers. The studies cover different groups and use different measures.
We should not exaggerate what the evidence tells us. But we should take the learning question seriously. Apprenticeships and placements need to provide more than a line on a CV: they need paid opportunities to practise, time with experienced colleagues and a clear route towards more demanding work. Simulations can help, but should complement real work experience rather than simply replace it.
What should VET teach?
The ETF report argues that AI literacy should become a foundational capability. That does not mean everyone needs to become an AI specialist, or that learning to write prompts is enough. Learners need to understand the limitations of AI, check its outputs and recognise issues such as bias, privacy and accountability.
This learning needs to be connected to particular occupations. A trainee vehicle technician might use an AI-supported diagnostic tool, but should also know how to check its suggestion against measurements and technical information. A business administration learner might use AI to draft a letter, then check it for errors and inappropriate use of personal data.
Crucially, judgement depends on knowledge. It is hard to spot a plausible but incorrect answer if you do not understand the work. VET must therefore continue to develop literacy, numeracy, occupational knowledge, practical skills and the ability to learn with others. Learners may benefit from practising both with and without AI, so that using the tool strengthens rather than bypasses their understanding.
Assessment must change too. A polished final product may reveal little about what a learner knows if AI has played a major part in producing it. Practical demonstrations, discussion and workplace projects can help learners show how they reached a decision, checked a result and applied their knowledge.
A shared responsibility
Teachers and workplace trainers need time, support and opportunities to learn about AI in their own fields. Colleges and training providers also need clear guidance on privacy, access and responsible use. Without investment, the gap between well-resourced and poorly resourced institutions could grow.
The wider issues go beyond education. AI can improve work, but it can also increase monitoring, intensify workloads or shift responsibility for mistakes onto workers. Training alone will not decide which outcome prevails. Workers, learners, employers and educators need a voice in how technology is introduced and how its benefits are shared.
The challenge for VET is not simply to prepare people to use new tools. It is to give them the knowledge, experience and confidence to work with them critically—and to help shape the future of their occupations and their society.
References
Adăscăliței, D. (2026, September 16). AI is already in use in Europe’s workplaces, but not at the expense of young workers. Eurofound.
Cedefop. (2025). Skills empower workers in the AI revolution: First findings from Cedefop’s AI skills survey (Policy brief). Publications Office of the European Union.
https://doi.org/10.2801/6372704
European Training Foundation. (2026). Changing landscape of skills in the age of AI. Publications Office of the European Union.
https://doi.org/10.2816/7083675
Palladino, M. G., Bergeaud, A., Bertheau, A., Bunel, S., & Degos, O. (2026, October 8). Generative AI: The first changes in the labour market are affecting young people [Blog post]. Banque de France.
About the Image
AI companion apps offer us intimacy on demand, free from the burdens of human relationships and tailored to our exact preferences. This image captures the unnatural process of constructing a partner in the same way we might customise a product. It depicts fragments of classical sculptures — symbols of idealised beauty — being digitally selected and rearranged into a perfect ‘soulmate’. But the result is merely a flawed collection of parts chosen for their individual appeal, not a real person.
