AI use in French companies: Insee 2025 figures by size and sector

In short: according to Insee Première no. 2120, published on July 21, 2026 (Clément Lefebvre, 2025 ICT survey of companies), 18% of companies located in France with 10 or more employees report using at least one AI technology in 2025, against 10% in 2024 and 6% in 2023. The rate tripled in two years and is now close to the EU average (20%). But the gap between small and large companies is widening: 15% below 50 employees, 58% at 250 employees or more.

Correction, September 23, 2026: the first version of this article mistakenly presented the 2024 rates by company size (9%, 15%, 33%) as 2025 figures. The figures below come from figure 1 of the Insee publication.

AI use by company size, 2023–2025

The catch-up is real: AI use tripled between 2023 and 2025 in companies with fewer than 250 employees. But because they started lower, the gap in percentage points grew — from 16 points between companies under 50 and those at 250+ in 2023, to 43 points in 2025. Insee attributes the size effect to the fixed costs of adoption. Companies under 50 employees make up 85% of the survey scope, which is why they pull the average down. Against the rest of Europe, French companies with 50 to 249 employees are now at the EU level, those with 250+ are 3 points above it, and those under 50 remain 2 points below.

By sector: from 9% to 59%

The raw rate understates some sectors. Weighted by employment, 66% of jobs in transportation and warehousing are in a company that uses AI, although only 9% of the sector’s companies do: the large groups have adopted it. Companies using AI now account for 66% of revenue and 59% of employment in the survey scope, against 49% and 40% in 2024 — which, Insee points out, does not mean that 59% of employees use AI themselves.

Why companies hold back: the stated barriers

Companies that already use AI name the same brakes on going further: 53% cite a lack of expertise, 43% data protection, 42% a lack of legal clarity and 33% costs. Use also remains narrow: among AI users, 56% use automatic text mining and 43% generative AI for text or speech, and 73% say they use AI to cut costs (saving time or labor, automating repetitive tasks).

What the research adds

Working on the same French ICT survey (2018 data), Flavio Calvino and Luca Fontanelli (Research Policy, 2026) separate companies that buy AI from a provider from those that develop it. Buyers are larger than other firms, but the difference is explained by their digital equipment: other digital technologies, digital skills and infrastructure are what matter. AI users are more productive, but mostly because already productive, digitally intensive firms are the ones that adopt AI — except for firms developing their own AI, where the link with productivity remains significant.

A systematic review by Leon Oldemeyer, Andreas Jede and Frank Teuteberg (Management Review Quarterly, 2024) lists 27 challenges faced by SMEs adopting AI; the most common are lack of knowledge, costs and inadequate infrastructure — the same ones that come out of the Insee survey.

For an SME that hasn’t started: where to begin

• →The first barrier small companies report is not cost but “no use” (71%). Start from one precise, repetitive, measurable task — sorting incoming email, chasing unpaid invoices, summarizing documents — not from a tool.
• →The second is expertise (53%). Plan for someone, internal or external, who supervises the first use case and checks its outputs.
• →Deal early with data protection and the legal frame, which more than a third of companies cite: what data the tool sees, where it is hosted, who can access it. See our article on the CNIL note on agentic AI.
• →Measure one indicator before and after (time spent, response time, error rate) over a few weeks before extending.
• →Digital equipment matters more than the AI tool itself (Calvino and Fontanelli): clean data and software that talk to each other are often the real first project.