Contents7
In short
Three serious public sources give, in the same year, 18%, 40% and 73% of companies using AI. None is wrong: they do not ask the same question of the same companies.
What changes everything: the definition of use (organised, declared, even very rare), the size of the companies surveyed, and what is counted (companies, employees or executives).
Size widens the gap: in France, organised use of AI ranges from 15% of companies with fewer than 50 employees to 58% of those with 250 or more.
Before quoting an adoption rate, or comparing yourself with one, five questions are enough. They are at the end of this article.
Three figures for a single question
A manager who searches for “how many companies use AI” finds three public answers within minutes, published a few months apart. All three come from France or the euro area, which makes them easy to compare and easy to confuse.
of French companies with 10 or more employees made organised use of AI in 2025, according to INSEE, the French statistics office
of French micro, small and medium businesses use an AI tool in 2026, according to the France Num barometer
of euro area companies use AI, even very rarely, at the end of 2025, according to the ECB survey
Side by side, these figures seem to contradict each other. They do not: each answers a different question. The trouble starts when they are compared with one another, or when a sales pitch picks whichever suits it.
What each one counts
INSEE: organised use, in companies with 10 or more employees
INSEE, the French public statistics office, surveys about 11,000 companies with at least 10 people, excluding agriculture, finance and insurance, as part of a European survey. It counts those that use at least one AI technology in their operations: text analysis, content generation, speech recognition, machine learning. It does not count the employee who opens an AI assistant on their own phone: individual, occasional use is outside the figure.
It is the most demanding definition, and the one that compares across countries: Eurostat, which publishes the same survey for the whole Union, finds 19.95% for the EU and 18.16% for France in 2025.
Watch out
59% of employment does not mean 59% of employees
INSEE notes that companies using AI account for 59% of employment in its scope. That does not mean 59% of employees use it: a 300-person company counts in full as soon as one department uses AI.
France Num: any tool, in companies with fewer than 250 employees
The France Num barometer, a public programme of the French Directorate General for Enterprise, surveys 9,655 companies with fewer than 250 employees, most of them very small. It counts any AI tool, including an assistant used without a paid subscription. Hence a higher figure. The same barometer gives a useful benchmark: 19% of these businesses pay for an AI tool.
The ECB: any use, even very rare
The European Central Bank’s survey of 5,067 euro area companies, 92% of them with fewer than 250 employees, asks how often the company uses AI. 27% answer never, 33% very rarely, 31% moderately and 7% significantly. The 73% therefore adds up everyone except those who never use it. The telling figure is elsewhere: significant use remains rare.
The Banque de France, which reread the same survey for France, keeps only moderate or significant use: 23% of French companies, against 39% in the euro area. The authors themselves point out that these data are self-reported.
Why size changes everything
A national average mixes the 12-person company with the 2,000-person one. And the gap between them is growing: according to INSEE, it went from 16 to 43 points between 2023 and 2025, which the institute attributes to the “fixed costs of adoption”. Putting AI in place takes a minimum of time, skills and tools whatever the size, and that minimum weighs more heavily on a small organisation.
Organised use of AI in France, by size, in 2025
For teams of 20 to 500 people, no official size class fits exactly: read the 10 to 49 and 50 to 249 bands as published, without merging them. INSEE adds a useful detail: French companies with fewer than 50 employees remain 2 points below the European rate, while those with 50 to 249 have caught up.
The traps that keep coming back
Comparing two definitions
The most common trap: writing that adoption “rose from 18% to 40%”. These are not two successive measurements, but two different questions put to two different populations. The same caution applies to barriers: INSEE asks all companies without AI (71% say they see no use for it), Eurostat only the companies that considered AI without adopting it (70% cite a lack of expertise). The two percentages do not share a denominator.
Confusing use with purpose
Using AI does not say what for. INSEE notes that a quarter of French companies using it stick mainly to generative AI, and that 28% of this group report no specific purpose. An adoption rate measures presence, not a use that changes how people work.
Taking a survey of large groups for an average
Consulting firm surveys question other companies. Deloitte, which sells consulting, audit and AI services, surveyed for its State of AI in the Enterprise 2026 3,235 executives from organisations it describes as on the leading edge of AI, all of which use it every day: an adoption rate makes no sense there, since it is 100% by design.
From the field
What our own register taught us
Before writing about business AI, we keep a register of figures published by consulting firms and official statistics. Closed on 14 October 2026, it lists 489 figures, of which only 278 were read at the source and judged ready to quote. The first rule we drew from it fits in one line: say, in the sentence itself, what an adoption rate counts.
Five questions before quoting an adoption rate
- Who was surveyed?
Companies, executives or employees, and from what size. A rate calculated on companies with 10 or more employees says nothing about sole traders.
- What use is counted?
Organised, declared, paid, or even very rare. This is the question that explains the gap between 18% and 73%.
- Who pays for the survey?
A public institute, a public bank that also supports these companies, or a firm that sells the consulting it recommends.
- Which fieldwork year?
A survey published in 2026 may describe 2025, and the figures move fast.
- What denominator?
All companies, or only those already using AI. A percentage without a denominator compares with nothing.
Our view
The right figure depends on the question. To know whether a company like yours has structured its use of AI, INSEE and Eurostat are the reference. To know how many teams already use it without management having organised it, France Num and the ECB say something more accurate. None of these figures says whether AI pays off: that is another question, covered in our article on AI return on investment.
And none replaces measuring your own company: knowing who uses what, for which task, and with what data. That is where integrating AI into a business begins, and it is often where unsupervised use comes to light, the subject of our article on AI governance.
To get a quick sense of where you stand, our digital maturity test gives a score across five areas. For a full, task-by-task assessment, that is the role of the AI Audit Sprint. And if AI is already running in your company and the bill is rising without you knowing what pays off, our AI at the right cost offer measures the cost of each use.
Frequently asked questions
What is the real percentage of companies that use AI?
There is no single one. For organised use in French companies with 10 or more employees, INSEE gives 18% in 2025; Eurostat gives 19.95% for the whole EU. For any tool in French companies with fewer than 250 employees, France Num gives 40% in 2026. Each figure is correct within its own definition.
Are small businesses lagging behind on AI?
They are below large companies, and the gap is growing: from 16 to 43 points between 2023 and 2025 in France, according to INSEE. But use also tripled in companies with fewer than 250 employees over the same period. The lag is mainly due to the fixed costs of getting started, which weigh more heavily on a small organisation.
Which European countries use AI the most?
On organised use, Eurostat puts Denmark (42.03%), Finland (37.82%) and Sweden (35.04%) ahead in 2025, with Belgium at 34.54%, Germany at 25.97% and France at 18.16%.
Why not use the big consulting firms’ surveys?
You can, as long as you say who they survey: usually companies with several hundred million dollars of revenue or more, and sometimes organisations that all already use AI. They tell you about large groups, not about average adoption.
Sources
- INSEE (new tab)insee.fr
- The France Num barometer (new tab)entreprendre.service-public.gouv.fr
- The ECB survey (new tab)ecb.europa.eu
- Eurostat (new tab)ec.europa.eu
- Banque de France (new tab)banque-france.fr
- State of AI in the Enterprise 2026 (new tab)deloitte.com
Jim · founder of Brain Plus AI
Also in business ai
Stepping back11 min
AI return on investment: why it is taking so long
The figure of 95% of AI projects with no return went round the world. What it really measures, why the other studies say roughly the same thing, and the practical lessons to draw from them for an SME.
Practical guide11 min The topic guide
Integrating AI in business: private and at the right cost
For teams of 20 to 500 people, integrating AI comes down to answering five questions in order: which task, where the data goes, which model, at what price, under what rules. This guide takes them one by one and points to our detailed articles.
Practical guide12 min
Enterprise AI governance: who can access what
An AI assistant reads whatever your access rights let it read, sharing mistakes included. Here is how to split your knowledge bases by department, keep a common core and decide who writes to the shared memory.