Europe’s AI productivity boost has turned up at last, and it is not evenly shared, according to a CEPR report.
Survey evidence covering more than 12,000 European firms suggests that adopting AI lifts labour productivity levels by about four per cent on average in the EU, with no sign of job cuts in the short term.
The report’s authors draw on the European Investment Bank Investment Survey, paired with Moody’s Orbis balance sheet data, to test what happens when firms actually deploy AI rather than just talk about it.
They show adoption rates that are similar on average between the EU and the US, but the details are a bit more awkward.
In financially developed EU countries such as Sweden and the Netherlands, about 36 per cent of firms reported using big data analytics and AI in 2024.
In less financially developed economies such as Romania and Bulgaria, adoption was around 28 per cent in 2024, and the gap has not narrowed.
Firm size matters just as much. About 45 per cent of large firms with more than 250 employees report deploying AI, versus 24 per cent of small firms with 10 to 49 staff.
AI adopters are also not typical firms. They invest more, innovate more, and are more likely to be squeezed by shortages of skilled workers, making simple comparisons unreliable.
To address causality, the authors use an instrumental variables approach that relies on matched US firms.
For each EU firm, they find comparable US firms by sector, size, investment intensity, innovation activity, financing structure and management practices.
They then use the adoption rate of matched US firms as a proxy for the EU firm’s exogenous exposure to AI, arguing that US institutional settings make this variation less intertwined with EU-specific factors.
The report found that AI adoption raises labour productivity levels by around four per cent, which the authors describe as economically meaningful but not the sort of boom that rewrites growth forecasts overnight.
They stress this is a one-off level effect in labour productivity, not a claim about long-run total factor productivity growth.
On jobs, the scary story does not show up. The report found no evidence that AI adoption reduces employment in the short run, even though a casual glance suggests AI adopters employ more people.
The mechanism they point to is capital deepening, where AI boosts worker output and decision-making without immediately displacing staff.
They also report that workers in AI-adopting firms have benefited from higher wages, both in aggregate and per employee, while noting that who retains those gains over the long term remains an open question.
Where the paper gets spiky is in distribution. Productivity gains are much stronger in medium and large firms than in smaller ones, which is a problem for Europe’s SME-heavy economy.
Additional investment in software and data infrastructure strengthens AI’s productivity effect; each additional percentage point of investment adds 2.4 percentage points to the effect.
Training is the bigger lever. An extra percentage point of investment in workforce training amplifies AI’s productivity gains by 5.9 percentage points.
That leads to policy recommendations that are less about handing out AI licences and more about making firms capable of using the stuff properly.
The report’s authors argue Europe needs better channels for capital to reach innovative, fast-growing, smaller firms so they can scale, linking this to progress on the EU Savings and Investment Union.
