ECB research finds jobs vulnerable to AI substitution declined between 2019 and 2025, while firms adopting the technology are actually more likely to hire
Between 2019 and 2025, positions at high risk of AI substitution declined in the US, while roles at low risk of replacement grew.
What the ECB actually found #
The European Central Bank’s research, spanning analyses published in 2025 and into early 2026, finds that firms utilizing AI significantly are approximately 4% more likely to expand their workforce, not shrink it.
On the productivity side, AI adoption has been observed to increase labor productivity by an average of 4% across the EU. That boost is particularly pronounced in sectors heavily invested in research and development.
The ECB’s data comes partly from the Survey on the Access to Finance of Enterprises, known as SAFE, which collects firm-level information on how businesses incorporate AI into their operations. The survey results indicate that AI use has been broadly neutral for employment in the euro area so far. High-intensity AI use actually displays a positive impact on employment, especially within R&D-heavy sectors.
The US picture is more complicated #
American labor market data shows declines in early-career roles within highly AI-exposed occupations, particularly after 2022 and 2023, when generative AI tools like ChatGPT went mainstream.
The pattern emerging from the US data is that jobs at high risk of AI substitution declined between 2019 and 2025, while jobs at low risk increased. The squeeze on early-career positions is particularly concerning because those roles traditionally serve as on-ramps into professional careers.
What this means for investors #
The 4% productivity gain associated with AI adoption is significant. For context, annual productivity growth in developed economies has hovered around 1-2% for much of the past decade.
The ECB’s data suggests firms integrating AI are also more likely to hire, meaning the productivity gains aren’t coming purely from headcount reduction.
The risk, which the ECB acknowledges, is that longer-term effects remain uncertain. The current data captures what amounts to the first few innings of AI integration. The US evidence on declining early-career roles in exposed occupations is a leading indicator worth monitoring.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our