France's AI Adoption Triples to 18% of Firms

Nearly one in five French companies with at least 10 employees — 18% exactly — used at least one artificial intelligence technology in 2025, according to an Insee survey published in late July. That is three times the level recorded in 2023, implying a base of roughly 6% two years earlier, and leaves France only slightly below the European Union average.

The result is striking for a country often described as a follower on new technology. Adoption remains closely tied to company size, with Insee attributing the gap to the fixed costs of introducing AI. The survey distinguishes eight use cases, including text analysis, data analysis, spoken-language or video generation, and decision support, applied in areas such as cybersecurity, organisation, accounting and logistics.

What the survey does not show is how many jobs will be destroyed, changed or created. Alexandra Roulet, an economics professor at Insead, estimates that a consensus between pessimists and optimists will not emerge for another five years. For now, the jobs question remains open even as enterprise AI use accelerates.

Why the Insee AI Numbers Matter for French Employment

Why Company Size Still Shapes Adoption

Insee points to fixed adoption costs as the main brake on smaller firms. AI tools require procurement, integration, training and governance, and larger employers can spread those costs across a bigger workforce. That explains why usage remains closely linked to company size: smaller French companies are more likely to wait for cheaper, more standardised options before adopting.

Advertisement

What the Eight Use Cases Tell Us About the Wave

The eight use cases named by Insee — text analysis, data analysis, spoken-language or video generation, and decision support — are not experimental. They appear in security, organisation, accounting and logistics, functions with repetitive, information-heavy tasks. This suggests the tripling is driven more by practical back-office deployment than by attention-grabbing projects.

Why the Jobs Question Will Not Be Settled Quickly

Alexandra Roulet's five-year horizon is not a forecast of mass layoffs; it is a statement about evidence. AI can destroy some tasks, change others and create new roles at the same time, so the net effect becomes visible only after several years of company-level data. French employers and workers should therefore treat the 18% adoption rate as an early indicator of structural change, not as a verdict on total employment.

What the 2025 Insee Survey Means for French Employers and Workers

The survey provides a measurable baseline rather than a precise jobs forecast.

  • Benchmark your firm against the 18%. The Insee figure covers companies with at least 10 employees. If your organisation is not among the 18% in 2025, the adoption gap is now quantified against both the national and EU averages.
  • Start with the eight use cases. Insee names text and data analysis, spoken-language and video generation, and decision support as the main applications, deployed in security, organisation, accounting and logistics. These are the clearest places to test how AI changes work in your own processes.
  • For employees, map exposure to those use cases. Roles built heavily on text analysis, data analysis or routine decision support are likely to feel change before the aggregate employment effect is settled.
  • Do not turn the 2030 horizon into certainty. Roulet's five-year consensus window means no reliable aggregate figure yet exists for destroyed, modified or created jobs. Workforce decisions should be based on actual task-level pilots, not on broad predictions.