ECB study flags limited impact of AI boom on US jobs and wages

A surge in artificial intelligence adoption may be displacing some workers, but its overall effect on aggregate employment and wages in the United States has remained limited so far, according to a study released by the European Central Bank on Monday.

The study comes as companies continue to invest heavily in AI tools, fuelling concerns that automation could replace workers at a growing pace, weaken hiring, and widen inequality.

While the debate over AI’s impact on labour markets has intensified globally, the ECB said the broader effects on total employment have so far been muted, even as signs of disruption are emerging in specific job categories.

Aggregate impact remains limited

The ECB said AI’s effect on labour markets is still difficult to measure at this stage, but recent evidence suggests the technology has not yet caused a broad decline in overall employment in the United States.

The study focused on the US labour market, where the effects of AI were expected to appear earlier than in other major economies.

The ECB said this was because the United States is home to some of the most advanced early adopters of AI and has a relatively flexible labour market.

The report noted that AI can affect employment in two opposite ways.

On one side, it can support job growth by boosting productivity and enabling firms to expand.

On the other hand, it can reduce employment by replacing workers in tasks that can be automated.

The final impact depends on which of these forces proves stronger.

Citing recent research, the ECB said firm-wide AI adoption in the United States has shown positive employment effects overall, but those gains have not been evenly distributed across occupations.

Early evidence from the European Union also points to higher productivity at firms adopting AI, without significant short-term labour replacement.

The ECB added that its own survey findings suggest companies with high AI adoption or AI-related investment are more likely to hire additional staff.

Pressure builds in high-risk occupations

Even though the aggregate picture remains relatively stable, the ECB study found growing signs of strain in occupations with a high risk of AI substitution.

Using an occupation-level index that groups jobs into low, medium, and high AI substitution risk categories, the study found that employment in high-risk occupations in the United States fell by more than 4% on average between 2019 and 2025.

These roles included occupations such as economists and graphic designers.

By contrast, employment in low-risk occupations, including electricians and high school teachers, rose by 13% over the same period.

This shift has changed the composition of the US labour market, the study said.

The share of low-risk jobs in total US employment rose to 25% from 23%, while the share of high-risk jobs fell to 33% from 35%.

AI linked to job reallocation

The ECB said its analysis shows that AI has already contributed to a reallocation of jobs within the US labour market.

To assess this, the study estimated how employment growth differed across occupations with varying levels of AI substitution risk between 2020 and 2025, using 2019 as the base year.

The analysis controlled for factors including sector-specific developments, the effects of the COVID-19 shock, and other unobserved differences across industries.

The findings showed an increasingly wide gap between employment growth in high-risk and low-risk occupations.

According to the study, jobs with a high risk of AI substitution grew by around 15 percentage points less than jobs with a low risk between 2019 and 2025, all else being equal.

Impact appears to be accelerating

The ECB said the results are consistent with a growing body of research showing that AI is affecting employment outcomes for specific occupational groups, particularly those more exposed to automation.

While the broader impact of AI on aggregate employment remains inconclusive, the study said the technology has already had a measurable relative effect on employment growth across occupations in the United States since 2019.

That trend appears to have accelerated since the launch of ChatGPT in late 2022, the ECB said, suggesting that AI’s labour market effects may become more visible as adoption deepens across industries.