Artificial intelligence is changing where Americans work, but it has not yet caused the mass job losses some forecasters warned about. A European Central Bank study released on June 22, 2026, found that U.S. employment in high-risk occupations fell between 2019 and 2025, while low-risk jobs grew. The overall effect on total employment and wages, however, remained muted.
The ECB Economic Bulletin article, titled “AI and the US labour market: effects on employment growth,” was written by ECB economists Isabella Moder and Til Pommer. It uses a difference-in-difference approach with an AI substitution risk index developed by Pizzinelli et al. (2023) at the International Monetary Fund. The study controls for sector-specific developments and base-year effects, with 2019 as the starting point.
Which jobs are losing ground to AI
The study splits occupations by how easily AI can substitute for the work performed. High-risk roles include economists and graphic designers. Low-risk roles include electricians and high school teachers. Between 2019 and 2025, employment in high-risk occupations declined by more than 4 percent. Employment in low-risk occupations rose by 13 percent.
| Job category | Examples | Employment change, 2019–2025 |
|---|---|---|
| High AI substitution risk | Economists, graphic designers | Down more than 4% |
| Low AI substitution risk | Electricians, high school teachers | Up 13% |
| Share of low-risk jobs in total U.S. employment | Across all occupations | Rose from 23% to 25% |
| Share of high-risk jobs in total U.S. employment | Across all occupations | Fell from 35% to 33% |
The gap is large. The ECB report states that, all else being equal, high-risk jobs grew by about 15 percentage points less than low-risk jobs over the six-year period. The shift became more pronounced after the launch of ChatGPT in late 2022.
Why the overall labor market has stayed stable
Despite the occupational shift, the U.S. economy has not seen a spike in unemployment or a collapse in wages. The ECB notes that the effects are likely to appear earlier in the United States than elsewhere because American companies are among the earliest adopters and the U.S. labor market is relatively flexible.
The study also found that AI exposure has “not yet translated into significant differences in wage growth.” The report cautions that income effects could become more pronounced over time. That would happen as the labor market keeps adjusting and AI tools become more generative.
What business leaders are saying about AI returns
The labor market data arrives as corporate leaders question whether AI investments are paying off. The PwC 28th Annual Global CEO Survey was released January 19, 2026. It found that 56 percent of CEOs worldwide said their companies had seen neither higher revenues nor lower costs from AI. Only 12 percent reported both revenue gains and cost savings.
Less than one-third of CEOs said their companies achieved additional revenues from AI over the previous 12 months. About one-quarter reported lower costs after implementation.
Two very different forecasts
Opinion remains split on the long-term outlook. Geoffrey Hinton, the computer scientist often called the “godfather of AI,” warned in an August interview that he is “fairly confident” AI will drive massive unemployment. He has also warned about the broader risk of developing AI systems that exceed human intelligence.
Elon Musk takes a more bullish view. In a May 19 interview with Forbes, Musk predicted that digital intelligence would exceed the sum of all human intelligence by 2031. He also said there could be 100 million to one billion humanoid robots within five years, and that the economy could double in size within five to six years.
What this means for investors
For investors, the ECB study offers a measured signal. AI is reallocating labor across occupations rather than eliminating work on a large scale, at least for now. That supports the case that AI adoption may be a longer productivity story than a near-term cost-cutting wave. The muted wage and employment effects also help explain why corporate AI returns have been uneven.
The data do not settle the debate between AI optimists and pessimists. They do show that the labor market adjustment is already visible in the occupation mix, even if aggregate unemployment remains low.
Stay ahead with our weekly newsletter
Want weekly analysis on AI, markets, and the economy? Sign up for the AlphaBetaStock newsletter and get the stories that matter delivered to your inbox.
Sources: European Central Bank Economic Bulletin, Issue 4, 2026; PwC 28th Annual Global CEO Survey; ZeroHedge / The Epoch Times.
