FRANKFURT, June 22 - A surge in corporate investment in artificial intelligence has raised concerns that machines could replace many workers, depressing employment and amplifying income inequality. Yet a study by the European Central Bank, published in its Economic Bulletin on Monday, finds that the aggregate effects on U.S. employment and wages have been limited so far.
The ECB's analysis notes that firms have indeed increased AI spending in recent years, but the U.S. labour market appears to have been adjusting for some time. Jobs judged to be most exposed to substitution by AI have been partly reallocated to other parts of the economy, producing a gradual reshaping of work patterns rather than a broad collapse in employment.
Quantitatively, the ECB highlights a sizeable divergence in growth between jobs with differing substitution risk. "All else being equal, between 2019 and 2025 jobs with a high substitution risk grew by around 15 percentage points less than jobs with a low substitution risk," the study said.
Employment in occupations identified as having a high risk of AI substitution, such as economists or graphic designers, fell on average by more than 4% between 2019 and 2025. By contrast, employment in occupations with a low risk of substitution, including electricians and high school teachers, rose by 13% over the same period.
The relative composition of U.S. employment changed as a result. The share of low-risk jobs in total employment increased from 23% to 25%, while the share of high-risk jobs decreased from 35% to 33%.
On incomes, the ECB found no measurable overall effect to date. "AI substitution risk has had no significant impact on wage growth since 2019," the institution said, while also cautioning that the situation could change. "Over time, as the labour market continues to adjust and AI tools become more generative, income effects may be more pronounced."
The study also points to heterogeneous impacts within sectors and workforce groups. Certain workers - notably junior staff in occupations that are highly exposed to AI - appear more vulnerable to displacement. Nevertheless, at the aggregate level through 2025, the data do not show a broad-based decline in employment or wages attributable to AI adoption.
Overall, the ECB frames the U.S. experience as one of gradual reallocation: roles with higher substitution risk have seen weaker growth or modest declines, while lower-risk roles expanded, leaving aggregate employment and wage trends largely unchanged so far. The study leaves open the possibility that further advances in generative AI and continuing labour-market adjustments could amplify distributional or income effects in the future.