The European Training Foundation (ETF), together with Cedefop, Eurofound, the European Commission, the International Labour Organization (ILO) and UNESCO, released its report Changing Landscape of Skills in the Age of AI (August 2026).
Human Level’s Take:
- Artificial intelligence is becoming increasingly embedded in the workplace, while the age-old fear that technology will take or replace jobs has returned with new urgency. ETF, however, points to a more nuanced reality. The emerging story is not simply one of jobs being lost to AI, but of work itself being reorganised around it, changing which tasks people perform, the skills they draw on, and where human contribution sits alongside increasingly capable systems.
- What emerges, then, is less a picture of machines simply replacing human work and more one of work being redistributed around new capabilities. As AI takes on more routine activities, workers are increasingly being shifted towards more complex tasks that depend on human judgement and context.
- This transition, however, is unlikely to be experienced evenly. ETF’s findings point to the possibility of deeper labour-market polarisation, with demand growing for highly skilled roles while some middle-skilled, white-collar jobs face greater exposure where tasks can be automated. At the same time, those most exposed to disruption are not necessarily those best positioned to adapt to it.
- This is where the training gap becomes particularly important. If AI is changing the mix of skills required across occupations, the question for businesses is not simply how many people receive AI training, but who receives it. ETF finds that low-skilled workers, often among those more exposed to automation, are also the least likely to have access to AI literacy training.
- For businesses, this shifts the challenge from simply rolling out AI tools or offering standalone AI training to thinking more deliberately about how work, skills and workforce development evolve together. What’s needed now is what ETF describes as “hybrid intelligence”, meaning that companies preparing for AI need to build technical literacy alongside domain expertise, judgement, adaptability, communication and continuous learning across their workforce.
Some key takeaways:
- AI is redefining work itself, not just individual tasks: The report finds that AI adoption is reshaping how jobs are structured, not only which skills are needed within them. Surveys show that around one-third of European workers no longer perform some tasks they did before adopting AI, while four in ten now do new or different tasks. ETF points to a parallel OECD study across finance and manufacturing which found a similar pattern. 66% and 72% of employers respectively said AI had automated tasks previously done by workers, while around half in each sector said it had also created new tasks that didn't exist before. The report describes this as going beyond simple automation toward the redefinition of whole work processes, with workers increasingly reallocated to more complex tasks as simpler ones are automated. This is reinforcing job polarization, where demand is rising for high-skill roles (and some low-skill services), while middle-skill white-collar jobs exposed to automation and offering little complementary human value face a higher displacement risk. Looking ahead, the report concludes that most human skills will still be needed, but how people use them will change as they work alongside AI systems, ultimately pointing to a shift in skills policy toward what ETF calls "hybrid intelligence," which combines technical AI literacy with domain expertise and distinctly human capabilities.
- A different set of skills is demanded from workers across the board: As jobs change, so too does the mix of skills people need to perform them. ETF finds that AI adoption is altering how individuals use cognitive, socio-emotional and physical abilities across a broad range of occupations, reshaping both the variety and depth of skills expected from workers. There is growing demand for higher-order cognitive skills, including analytical ability, critical thinking and problem-solving, alongside a parallel rise in the value of socio-emotional skills. Tasks involving empathy, creativity and leadership carry just a 13% potential for AI transformation, since they depend on human judgement and context. At the same time, simple, routine manual tasks are becoming less central as AI-enabled automation takes on a greater share of them. Digital skills are also deepening in both variety and depth: more than 8 in 10 EU jobs (87%) now require at least basic digital skills, and 68% of digital skills are expected to change in how they are applied as AI reshapes workflows. Human strengths such as adaptability, ethical judgement, cross-disciplinary communication and continuous learning are, according to the report, becoming more valuable rather than less, as human-AI collaboration increasingly depends on people providing framing, oversight and context that AI cannot supply on its own.
- Businesses are training too little and too unevenly to keep pace with the skills shift: Despite these changing demands, ETF finds that talent and skills gaps are widening for many firms, especially in emerging AI-related roles where qualified candidates remain scarce. Yet AI literacy remains low even in advanced economies, with only around 15% of European employees taking part in AI-related training in 2023–24. Some employers are moving to close that gap, with public administration, finance and health among the sectors introducing AI awareness courses, and programmes such as IBM SkillsBuild, Google AI and PwC’s Digital Fitness cited as employer-led examples. But the report describes these as a useful start rather than a fix. System-wide investment is still missing, and there is not yet a reliable way to recognise skills built outside formal training. A sharper blind spot emerges for low-skilled workers: often among those most exposed to automation, they are also the least likely to have access to AI literacy training. ETF highlights that closing this gap will require more inclusive pathways and targeted incentives. Looking ahead, the challenge is not simply to train more people in AI, but to develop the broader combination of technical knowledge, domain expertise and human capabilities needed to work effectively alongside it, while shifting from certifying fixed tasks towards recognising people’s capacity to learn and adapt.
