Mind the Gap: AI Adoption in Europe and the U.S. (w34995)
S · Shapes the field Brookings Papers on Economic Activity paper by Bick, Blandin, Deming, Fuchs-Schündeln and Jessen, the team behind the most cited US AI-adoption surveys, with new comparable worker surveys in seven countries and the first academic use of the EU firm AI module.
Summary based on the full text·Original: about 90 minutes to read
In short
NBER Working Paper 34995 (March 2026), to be published in the Spring 2026 Brookings Papers. Worker AI adoption is clearly higher in the US than in Europe, and varies widely inside Europe, from 36% in the UK to 26% in Italy. About half the gap comes from differences in occupations, industries and firm size; most of the rest lines up with whether employers manage people well and actively encourage AI use. Industries with more AI adoption have had faster productivity growth, with no clear change in employment.
Key numbers
- 43%US workers using AI for their job in 2026, against 32% of European workers surveyed
- 7%US firms using AI for production in 2025, against 4% of EU firms
- 55%share of the adoption gap with the US explained by composition (occupation, industry, firm size, demographics)
- 2.3%aggregate work-time savings from AI among all US workers, against 1.0% to 1.8% in the European countries surveyed
- 2 - 5 percentage pointsextra cumulative productivity growth associated with 10 points more AI adoption in European industries (not causal)
As reported by the source.
English·Web page·opens nber.org
What it reports
- 43% of US workers use AI for their job versus 32% in the European countries surveyed. 7% of US firms use AI for production versus 4% of EU firms. In Europe, firm use ranges from 9% in Sweden to under 1% in Serbia.
- Counting all business uses, European firms use AI about five times more than the "production" measure shows. The authors argue the US Census BTOS still undercounts firm adoption.
- European firm adoption was flat from 2021 to 2023 and has more than doubled since, with leaders pulling ahead.
- An Oaxaca-Blinder decomposition attributes 55% of the gap to composition. Adding employer encouragement of AI accounts for nearly the whole US-Europe gap. Encouragement and tool provision predict use; training does not.
- AI time savings imply 2.3% aggregate time saved in the US and 1.0% to 1.8% in Europe, worth 0.5 to 1.3 points of US labour productivity advantage.
- Across 29 European countries, 10 points more industry AI adoption is linked to 2 to 5 points more productivity growth to 2024. US estimates are similar: 3.7 points for 2019-2025. A 2015-2019 placebo shows no effect.
- No clear link between industry AI adoption and employment change in either region.
How they got it
Two rounds of nationally representative, comparable worker surveys in the US, Germany, UK, France, Italy, Sweden and the Netherlands: about 5,000 responses per country in May to June 2025 and about 3,000 per country in January to February 2026. Firm data from the US Census BTOS and the EU ICT Usage and E-Commerce in Enterprises survey (32 countries). Industry productivity regressions with country and industry fixed effects.
Limits
- The authors stress the productivity links are not causal and cover a short window.
- Time savings are self-reported by workers.
- The EU firm data are available only at country-industry level, not firm micro data.
Why it matters for companies
The strongest practical finding is that employer encouragement and tool provision explain adoption better than training does. That moves the lever from course catalogues to managers explicitly asking people to use AI and giving them sanctioned tools. For European firms, the gap with US peers is largely a management-practice gap that companies can close themselves, without waiting for regulation or better models.
Summary based on the full text. Read the original for the full argument and data.Read the original on nber.org
- Tags
- AI adoption, Europe