AI Impact on Employment: 4% More Hiring, but Entry-Level Jobs Shrink

The relationship between artificial intelligence and employment has become one of the defining economic questions of our time. New research from the European Central Bank cuts through the noise with some surprising findings: when it comes to the AI impact on employment, the story is more nuanced — and more hopeful in places — than the doom-and-gloom headlines suggest.

Key takeaways

• ECB research shows firms that adopt AI are approximately 4% more likely to expand their workforce, not reduce it.
• AI adoption is linked to a roughly 4% increase in labor productivity across the EU — double or more the typical annual productivity growth in developed economies.
• Jobs at high risk of AI substitution declined in the US between 2019 and 2025, while lower-risk roles grew.
• Early-career positions in heavily AI-exposed US occupations have contracted since 2022–2023, raising concerns about professional entry points.
• The ECB and independent research from Yale’s Budget Lab both conclude that long-term employment effects remain uncertain, and current data only captures the opening phase of AI integration.

ECB Research Finds AI Linked to Workforce Growth and Productivity Gains

Contrary to what many workers fear, firms integrating AI are not systematically cutting staff. ECB research finds those companies are actually about 4% more likely to increase their headcount than peers that haven’t adopted the technology. That single figure reframes a lot of the conversation around automation.

On the productivity side, the numbers are equally striking. AI adoption boosts labor productivity by roughly 4% on average across the EU — a meaningful jump when you consider that annual productivity growth in developed economies has typically hovered between 1% and 2% for much of the past decade. In other words, AI is delivering productivity gains at roughly double the historical baseline, at least in the early data.

That doesn’t mean every firm or every worker benefits equally. The ECB’s evidence draws partly from the Survey on the Access to Finance of Enterprises (SAFE), a firm-level dataset tracking how businesses incorporate AI into their operations. The SAFE data show AI use has been broadly neutral to positive for employment across the euro area — and in cases of high-intensity AI use, the employment impact tilts clearly positive.

Sector-Specific Productivity Boosts from AI

The productivity gains from AI are not distributed evenly. Research and development intensive sectors see the strongest improvements — a finding that makes intuitive sense, since AI tools that accelerate data analysis, hypothesis testing, and iterative design work are a natural fit for R&D-heavy environments.

This sector-specific concentration matters for how policymakers and investors should interpret the headline 4% figure. The EU-wide average likely masks sharper gains in sectors like pharmaceuticals, advanced manufacturing, and technology, alongside more modest effects in industries where AI applications are still shallow or nascent.

The SAFE survey findings reinforce this picture. Firms with the deepest AI integration are seeing employment grow, not contract. The narrative of AI as a pure job-killer simply doesn’t hold up against the current data — at least not in Europe’s business population.

US Employment Trends Show Mixed Impact of AI

The American labor market tells a more complicated story, and it deserves careful reading. Between 2019 and 2025, positions at high risk of AI substitution declined in the US, while jobs at lower risk grew. That structural shift is already underway — not something happening in a hypothetical future.

Decline in Early-Career Roles in AI-Exposed Occupations

The sharpest concern in the US data involves entry-level workers. Early-career roles within highly AI-exposed occupations have contracted, particularly after 2022 and 2023 — the period when generative AI tools like ChatGPT entered mainstream use. These positions traditionally function as professional on-ramps, the first rungs on career ladders in fields like finance, law, consulting, and tech. Their decline raises a structural question that goes beyond simple job counts: if fewer people can gain experience in AI-exposed fields at the entry level, who fills those roles a decade from now?

Independent research from Yale’s Budget Lab adds important context here. Their analysis found that AI has had a modest impact on America’s overall job market since ChatGPT’s release in 2022 — more comparable to the disruption caused by computers in the 1980s or the internet in the 1990s than to a seismic restructuring. Yale researchers put it plainly: AI usage shows “no connection” to changes in overall employment or unemployment rates. Occupational churn is following a trend line similar to previous technology transitions, not triggering a massive reset.

Growth of Jobs at Low Risk of AI Substitution

On the other side of the ledger, roles less exposed to AI automation have grown consistently across the 2019–2025 period. This bifurcation — high-exposure roles shrinking, low-exposure roles expanding — is a pattern worth watching. It suggests the labor market is already sorting itself around AI risk, even if the overall unemployment numbers haven’t spiked.

Sectors like finance and business services appear more vulnerable than occupations like nursing, where human judgment, physical presence, and interpersonal care remain difficult to automate. The Yale analysis found that high AI exposure doesn’t dramatically extend how long displaced workers remain unemployed — those out of work for under five weeks and those unemployed for 27 weeks or more show relatively similar trend lines. That’s a nuanced finding that complicates simple narratives in both directions.

What the Numbers Actually Mean

Stepping back, two findings from the ECB and corroborating research deserve emphasis for anyone trying to understand where this is heading.

First, the productivity numbers are genuinely significant. A 4% productivity lift associated with AI adoption — against a backdrop of 1–2% annual productivity growth in most developed economies — suggests AI is already moving the needle in measurable ways. The ECB’s data indicates these gains are not coming purely from headcount reduction; firms adopting AI are also more likely to hire. That points toward a dynamic where AI is augmenting workers rather than simply replacing them, at least in the current phase.

Second, the early-career employment signal in the US is the most important leading indicator to monitor. Entry-level declines in AI-exposed occupations don’t appear in headline unemployment figures, but they could compound over time. If fewer workers build foundational experience in AI-adjacent fields today, the skilled labor pipeline in those industries narrows downstream. That’s the kind of slow-moving structural risk that doesn’t trigger alarm bells immediately but shapes workforce composition over a generation.

The ECB is explicit about the limits of what current data can tell us. Long-term impacts of AI on employment remain genuinely uncertain. What exists now captures only the opening phase of AI integration across economies — a technology that is still evolving rapidly, with applications that were not possible even three years ago becoming standard practice today. The current data is informative, but it is not a forecast.

What emerges from combining ECB research with Yale’s findings is a picture that is neither the job apocalypse nor the pure productivity paradise that advocates and critics each tend to emphasize. AI is reshaping work — changing tasks, shifting which occupations grow, and concentrating early-career pressure in exposed fields — without yet producing the large-scale unemployment wave that some predicted. Whether the current equilibrium holds as AI capabilities accelerate into the late 2020s is the question neither dataset can yet answer.

FAQ

How does AI adoption affect workforce size according to ECB research?

ECB research finds firms using AI are approximately 4% more likely to expand their workforce rather than reduce it, suggesting that AI adoption tends to complement hiring rather than replace it outright.

What productivity improvements are linked to AI in the EU?

AI adoption increases labor productivity by roughly 4% on average across the EU, with stronger gains concentrated in research and development intensive sectors. For context, typical annual productivity growth in developed economies runs between 1% and 2%.

What employment trends have been observed in the US relating to AI?

Jobs at high risk of AI substitution declined in the US between 2019 and 2025, while lower-risk roles grew. Notably, early-career positions in highly AI-exposed occupations contracted particularly after 2022–2023, when generative AI tools became mainstream — a trend that raises concerns about professional entry points and long-term career pipeline effects.

Are the long-term employment effects of AI clear?

No. The European Central Bank explicitly acknowledges that while initial data show broadly neutral to positive effects in the euro area, the long-term impact of AI on employment remains uncertain. Current evidence covers only the first phase of AI integration, and the technology continues to evolve rapidly.

Article produced with the assistance of artificial intelligence and reviewed by the editorial team.