AI adoption in France: Insee’s new figures show how far SMEs are behind

On July 21, 2026, Insee published its annual survey on information and communication technologies in French companies, and for the first time gave a full-year read on AI adoption: 18% of companies established in France with 10 or more employees declared using at least one AI technology in 2025, up from 10% in 2024 — an 8-point jump in a single year, and roughly three times the 2023 level. That rate is now close to the EU average of 20%. It is genuinely fast growth. It is also an average that hides a much more uncomfortable number for most business owners.

An average that hides a gap by company size

Broken down by size, the picture changes: 9% of companies with fewer than 50 employees use AI, against 15% for those with 50 to 249 employees, and 33% for those with 250 or more — nearly four times the rate of the smallest companies. The gap shows up by sector too: usage stays at 9% in transportation and warehousing, 10% in construction, 12% in accommodation and food services, and 13% in retail, against 59% in the information-communication sector. Insee also notes that companies already using AI concentrate 66% of total business revenue and 59% of total employment in France — meaning the minority that adopted early already carries most of the economic weight.

A gap the research documents too, not just Insee

This is not a French peculiarity. A 2025 study by Serena Proietti and Roberto Magnani, Assessing AI Adoption and Digitalization in SMEs: A Framework for Implementation, examining small and medium-sized companies in Italy, finds the same structural pattern: a significant and persistent gap between SMEs and large corporations in AI use, driven less by a lack of interest than by concrete barriers — cost, the smaller and lower-quality datasets SMEs work with compared to large firms, and a shortage of in-house technical skills to evaluate and run a first project. In other words, the gap is not a temporary lag that will close on its own as tools get cheaper; it reflects the same three constraints in every SME regardless of country.

What it means for an SME that hasn’t started

If your company has fewer than 50 employees, a 9% adoption rate means you are still in the large majority, not lagging behind some private curve you should already be on. But the Insee figures also show the gap widening, not closing — companies already using AI keep compounding gains in revenue and headcount share, while the barriers Proietti and Magnani describe (cost, data, skills) do not resolve themselves by waiting. The practical response is not a broad "AI strategy," which is exactly the kind of undertaking those barriers make hard to start. It is picking one narrow, high-volume, low-risk process first.

• →Start with the use cases Insee shows growing fastest inside companies already using AI — automatic text analysis and generative AI for administrative tasks — rather than a customer-facing project: sorting and routing incoming email, or turning scattered figures into a standing report, are both lower-risk entry points than a chatbot.
• →Set one measurable indicator (hours saved, error rate, response time) over a 60-to-90-day pilot before deciding whether to extend it — the goal is evidence, not a company-wide rollout on day one.
• →Treat the cost and skills barriers as the real obstacle, not a nice-to-have: budget for someone (internal or external) who can supervise the pilot, since an unsupervised process is where the gap the research describes tends to widen fastest.

The number to keep in mind is not 18% or even 33% — it is that the gap between company sizes is growing, not shrinking, three years into the trend. Starting a small, supervised pilot now costs far less than closing a compounding gap later.