AI Creates More Jobs Than It Destroys in Europe Today: What the ECB Survey of 5,300 Companies Says

On March 4, 2026, the European Central Bank published an analysis based on a survey of 5,300 companies in the euro area. The central finding: companies that use AI intensively are 4% more likely to hire additional staff than those that do not. Only 15% of companies using AI cite labor cost reduction as their primary motive. These two figures contradict the widespread idea that AI is, in the short term, a net job destroyer in Europe.

This is not a partisan think tank study. It is the ECB, using a standard econometric methodology (ordered probit, controlling for size, age, sector, country, economic outlook), on data collected in 2025. The results deserve to be read precisely — neither as a definitive absolution of AI, nor as proof that employment concerns are unfounded.

The SAFE Survey: What it Measures and How

The SAFE (Survey on the Access to Finance of Enterprises) survey is regularly conducted by the ECB among companies in the euro area. For this analysis, Laura Lebastard and David Sondermann utilized data from two waves: Q2 2025 for AI investments, Q4 2025 for AI use. The final sample: approximately 5,300 companies.

The central question posed to companies was twofold: do they use AI? And if so, at what intensity and for what reasons? The responses allow for distinguishing three profiles: companies that do not use AI, those that use it occasionally, and those that use it intensively. It is this distinction that yields the most interesting results.

The Results: What the Data Exactly Says

The first surprising result is the gap between usage and investment: two-thirds of companies use AI, but only a quarter invest in it. This reflects the reality of generative AI accessible via online tools (ChatGPT, Copilot, Gemini): companies can benefit from AI without deploying significant capital.

The second result is the heterogeneity according to usage intensity. Overall, there is no significant difference between AI and non-AI companies regarding job creation or destruction. But when distinguishing intensive users from occasional users, the positive effect appears: +4% probability of hiring for intensive users.

The third result is the most nuanced: companies that use AI to reduce labor costs indeed have a negative effect on hiring and a positive effect on layoffs. AI can destroy jobs. But these companies represent only 15% of AI users.

Why AI-Intensive Companies Hire

The positive effect on employment is primarily driven by companies that use AI for R&D and innovation — not to automate existing tasks, but to develop new products, services, or processes. This is consistent with economic literature on general purpose technologies (GPT — General Purpose Technologies).

The effect is also driven by small businesses, not large ones. For large companies, AI has a neutral effect on employment in the short term. For small ones, it is positive. This result is counter-intuitive compared to the dominant narrative, but it is consistent with the idea that small businesses use AI to grow and reach a critical size they could not have achieved without this tool.

What the Study Doesn't Say

Three limitations deserve to be highlighted. The time horizon is short: the survey measures current hiring decisions and intentions for one year. The European results are not transferable to the United States, where adoption is more advanced. And correlation is not causation: companies that use AI intensively may already be growing for other reasons.

In Europe, 25% of Companies Use AI, but Only 5% Do So Intensively

Europe presents an AI adoption profile distinct from that of the United States and China. The European AI Act imposes transparency and compliance obligations. The OECD notes that Europe lags in investments but leads in regulation and ethical standards. In France, the AI Plan 2 for 2026-2030 foresees 5 billion euros in public investment.

The Workers Who Win and Those Who Lose

The aggregate positive effect hides a highly unequal distribution. The workers who gain the most are those with skills complementary to AI. The workers who lose the most are those whose tasks are directly automatable: data entry clerks, standardized customer service operators, reconciliation accountants.

The OECD estimates that 14% of jobs in developed countries are at high risk of automation (more than 70% of tasks automatable), and that 32% are at moderate risk. These figures do not mean that these jobs will disappear — they mean that the workers who hold them will have to adapt their skills.

What This Means for Public Policies

Vocational training is the priority lever. Social protection must adapt to sectoral transitions. AI regulation must be calibrated between innovation and protection. Social dialogue is essential.

The ECB study answers a precise question: in 2025-2026, in the euro area, does AI destroy more jobs than it creates? The answer is no — the net effect is slightly positive, driven by companies using AI for innovation rather than cost reduction. What is certain is that the debate on AI and employment can no longer take place without empirical data.