AI adoption raises labour productivity at European firms by an average of 4%, but the gains are concentrated among larger companies. Iñaki Aldasoro and colleagues report for the Centre for Economic Policy Research (CEPR) based on data from more than 12,000 European firms.
The researchers used a novel method to establish causation rather than mere correlation. They matched each European firm with comparable US companies sharing similar characteristics such as sector, size, and innovation activity. The AI adoption rates of those US firms then served as a proxy for how exposed their European counterparts would be to the technology under different conditions. This approach helps avoid the distortion that comes from the fact that firms which adopt AI are already more productive and innovative to begin with.
The central finding is clear: AI acts as a complementary tool that amplifies worker output without replacing jobs. The study finds no evidence of negative employment effects in the short run. Workers at AI-adopting firms also earn higher wages, though whether those gains will persist and be shared across skill levels remains an open question.
The productivity bump, while meaningful, is more modest than some optimistic forecasts have suggested. It reflects a one-off efficiency improvement rather than sustained long-run growth in total factor productivity.
The gap between winners and losers is striking. Large firms with more than 250 employees see substantially stronger productivity gains than small firms employing between 10 and 49 people. Adoption rates tell a similar story: 45% of large firms have deployed AI compared with 24% of small ones. The researchers point to scale as the key factor. Larger companies have the resources, technical expertise, and organisational capacity to absorb integration costs and make the complementary investments that AI requires.
Those complementary investments matter enormously. The study finds that each additional percentage point spent on software and data infrastructure amplifies AI’s productivity effect by 2.4 percentage points. Workforce training has an even larger impact: each additional percentage point invested in training boosts the productivity gain by 5.9 percentage points. Buying the technology is not enough. Firms need to redesign workflows and build what the authors call “fusion skills,” including prompt engineering and data stewardship.
Geography also shapes who benefits. Firms in financially developed EU countries such as Sweden and the Netherlands match US AI adoption rates, with around 36% of firms using AI. In less financially developed economies such as Romania and Bulgaria, the figure drops to around 28%.
The authors draw clear policy conclusions. Europe needs well-functioning capital markets to help smaller firms scale up and access finance. Public support should go beyond subsidising software licences and focus on training and integration. And while immediate job destruction appears unlikely, the authors warn that more capable AI systems could shift that picture over time. Monitoring labour market effects and ensuring inclusive growth will be essential.
