AI is gaining ground in French firms: rapid diffusion but productivity gains are so far limited
Bulletin No. 266, article 1. Artificial intelligence (AI) is spreading rapidly among French firms: at the start of 2026, more than two-thirds of firms with 20 or more employees reported using generative AI tools. This figure nonetheless masks wide disparities in use. Most firms have primarily adopted general-purpose AI tools, notably for support and cross-functional activities such as sales and administration. By contrast, more knowledge-intensive sectors use AI for specialised purposes and have integrated it more deeply into their production and innovation processes.
For the time being, the observed economic effects remain limited and consist primarily of productivity gains. However, firms expect more substantial benefits over the next three years, including greater productivity gains and a moderately negative impact on employment. They also plan to continue investing in AI and do not anticipate any significant fall in their selling prices. These results suggest that the main economic challenge now lies less in the adoption of AI itself than in firms’ ability to use it to deliver lasting improvements in their organisation and performance.
1 Two-thirds of French firms with 20 or more employees reported using generative AI at the start of 2026, mainly in support functions
Between the end of February and the start of April 2026, some 7,000 firms that take part in the Banque de France Monthly Business Survey were questioned on their use of artificial intelligence (AI). The findings were adjusted to be representative of the firm sizes and sectors covered by the survey, which together account for roughly half of the French economy. There was significant heterogeneity in AI adoption rates depending on firm size and type of AI.
Unlike non-generative (or predictive) AI, generative AI has a low entry cost and is spreading rapidly
Generative AI, which emerged more recently than predictive AI tools, has already been adopted widely (see Chart 1a). At the start of 2026, 67% of firms with 20 or more employees reported using it, mainly on an experimental basis or to a limited extent. Generative AI has a relatively low cost of entry, both in terms of technical skills and data requirements (it can operate using unstructured text). Employees can use it directly through standardised tools, such as chatbots (e.g. ChatGPT), or as part of software suites such as Microsoft Copilot. Adoption rises rapidly with firm size: 64% of the smaller firms in the survey (20 to 49 employees) reported using it, compared with 83% of larger firms (1,000 employees or more).
Non-generative, or predictive, AI is significantly less widespread (see Chart 1b): only around a third of firms with 20 or more employees reported using it in the survey. It often requires structured data and stronger integration capabilities, which tends to limit adoption to firms with a higher degree of digital maturity. Adoption of non-generative AI appears to be more sensitive to firm size: 31% of smaller firms (20 to 49 employees) reported using it, compared with 60% of larger firms (1,000 employees or more). Larger firms are also more likely to use it intensively, as they have a greater number of potential use cases, more financial resources and specialised skills enabling them to test and deploy AI solutions at scale.
The findings point to two distinct and complementary adoption dynamics: (i) widespread and rapid diffusion of generative AI, driven by the easy accessibility of these tools; and (ii) a more gradual diffusion of non-generative AI, which depends more on firms’ organisational and digital …
Download the full publication
Updated on the 3rd of September 2026