The occupational impact of AI in the generative AI era

AI is advancing faster in the lab than in the labour market, but its footprint across jobs is widening quickly.

To understand this footprint better, the Commission’s Joint Research Centre has published a new working paper called “Revisiting the occupational impact of AI in the generative AI era”. It traces how exposure to AI has evolved in the European labour market between 2008 and 2024. The analysis links 352 AI benchmarks to 14 cognitive abilities, 108 work tasks and 127 ISCO-3 occupations, weighting each benchmark by its research intensity in the AI literature. The result is a detailed map of how advances in AI translate into exposure for different kinds of work.

The most striking change concerns ideas-related cognitive abilities. Recent AI research increasingly targets capacities such as attention and search, comprehension and expression, and logical reasoning. These abilities underpin the kinds of information-processing and problem-solving tasks that are common across a wide range of occupations.

Because these tasks cut across professions, AI exposure has increased across the entire labour market. The study finds a sharp and sustained rise in exposure since the late 2010s, accelerating with the recent wave of generative AI systems.

Exposure is not evenly distributed. High-skill occupations remain the most exposed because they rely more heavily on cognitive and analytical tasks that current AI systems can support or automate. Jobs involving research, writing, analysis, coding and information management are particularly affected.

Yet the paper also finds that AI exposure is spreading beyond traditional knowledge professions. As AI capabilities expand, tasks previously considered less susceptible to automation are increasingly touched by AI tools. This creates what the authors describe as a transversal impact across occupational categories, including mid-skill and administrative roles.

Another key insight is that tasks matter more than job titles. Occupations are bundles of activities. AI systems typically affect only part of those activities. As a result, the likely outcome is not the disappearance of entire professions but a reconfiguration of task composition within jobs.

The findings challenge earlier assessments of automation risk that were developed before the emergence of generative AI. Systems capable of producing text, analysing information and assisting reasoning now target the very cognitive abilities that structure many forms of modern work.

For policymakers and organisations, the implication is straightforward. Understanding how AI interacts with specific tasks and cognitive abilities offers a more realistic picture of labour-market change than broad claims about which jobs will survive. The transformation underway is less about replacing workers than about reshaping how knowledge work itself is performed.