Contents7
French employees use AI, often without having learned how. They use it to write, search and summarise, sometimes in secret, and most have never received any training at all. Meanwhile, companies buy tools. The gap between what is spent on AI and what is invested in the people who use it has become the real barrier.
In short
According to KPMG, which sells audit, consulting and AI services, and the University of Melbourne, which surveyed more than 48,000 people in 47 countries, 24% of French respondents feel trained in AI: France ranks 40th out of 47.
INSEE, the French statistics office, finds that 53% of French companies that use AI are held back by a lack of expertise.
Accenture, which sells consulting and AI services, finds that 86% of companies plan to increase AI spending, but only 43% plan to train their teams.
A recent study suggests that heavy reliance on AI can hinder learning: training also means learning when not to delegate.
What the figures say about France
The KPMG and University of Melbourne study, funded by KPMG, is about people, not companies: more than 48,000 respondents in 47 countries, surveyed in late 2024 and early 2025 through representative panels. It measures what each person says about themselves.
of respondents in France feel trained in AI (40th out of 47 countries)
have received no training, against 61% worldwide
nonetheless use AI regularly
The same release adds two figures that change the reading: 57% of French employees surveyed hide their use of AI and present the content as their own, and 45% do not check the accuracy of results. The use is there, the learning is not.
Public statistics point the same way. Eurostat counts 18.44% of French people aged 16 to 74 who used generative AI for work in 2025 (Eurostat, isoc_ai_iaiu), and on the company side, INSEE notes that among companies with 10 or more people that use AI, 53% are held back by a lack of expertise. It is not the technology that is missing, it is the skills to use it.
Companies spend more than they train
The large international surveys reach the same conclusion. Accenture, which sells consulting, integration and managed AI services, surveyed 1,320 executives and 4,560 employees in 12 countries, without stating any company size threshold.
What companies surveyed by Accenture plan to do
PwC, which sells audit, consulting and AI services, surveyed 49,364 employees in 48 countries and territories: daily use of generative AI rose from 14% to 22% in a year, but only 51% have access to the training they need, down from 59% a year earlier. Use goes up, access to training goes down.
At Deloitte, which sells consulting, audit and AI services, the “State of AI in the Enterprise 2026” survey of 3,235 executives at organisations on the leading edge of AI ranks the skills gap as the top barrier to integration, yet only 53% of those organisations train all their employees.
Key takeaway
A tool is bought in a day, a skill is built over weeks
AI budgets are growing faster than training budgets, in large companies as elsewhere. That is where much of the expected value is lost.
Training, but in what?
AI training is not about learning to write prompts. A recent study raises a deeper question. In a working paper published in January 2026, Shen and Tamkin observe that using AI harms conceptual understanding, code reading and debugging, with no significant efficiency gain on average. It is a working paper, not a peer-reviewed article, and it is about programming: it does not say AI makes people incompetent. It says that heavy reliance can prevent people from learning what they delegate.
That risk is not a reason to keep AI from teams. It is a reason to train them to judge. We covered this debate in Does AI make us dumber? What the studies say and in The human role alongside AI: knowing how to judge.
What useful training teaches, in order
- The craft
What a good result must contain, before any tool.
- The tool
What it does well, what it does badly, what you never hand it.
- The review
Spotting the plausible error, checking a source.
- The rule
Which data may leave, who answers for the result.
Jobs do not become useless: their centre of gravity moves. An accountant who reviews the entries proposed by AI remains an accountant; she needs her expertise to spot what is wrong. A beginner who no longer does the simple tasks has to learn differently, which we discussed in Will AI take young people’s first jobs?.
What the law already asks
Since its rewording by Regulation (EU) 2026/1744, in force since 27 July 2026, Article 4 of the EU AI Act asks companies that use AI systems to take measures to foster AI literacy among their staff, without imposing a specific level for each person. It is not an exam, it is an approach you must be able to show. Our guide to the AI Act and GDPR for small businesses details what this obligation requires.
What we think
The 24% figure does not say French people are behind on AI: 67% already use it regularly. It says they use it alone. In a team of 20 to 500 people, that produces uneven practices, invisible to management, and sometimes risky when confidential data ends up in a consumer tool.
From the field
Reviewing is a skill, as a client learned the hard way
Our SEO engine, running since 25 August 2026, writes article drafts; none goes live without human review. That rule comes from a client who learned it the hard way: three pieces of content generated and published without review contained invented facts and had to be taken off the site. Reviewing AI text, checking every figure and every link, is a skill that can be learned.
Our position is simple: train before deploying, and train on the team’s real files rather than generic examples. Training that starts from the craft also reassures employees, because it shows that their expertise is the condition for using the tool well, not its victim.
Where to start
Start by finding out what is already happening: ask each team which tools it uses, for which tasks, with which data. Hidden uses come out when they are not punished. Then pick two or three tasks per role and train on those, review included. To frame access and data, our guide to enterprise AI governance sets the rules, and our guide to integrating AI in your business gives the overall method.
Our AI training by role offer starts from this principle: workshops on each team’s real files, then 30 days of coaching, for €4,900 for up to 15 participants.
Frequently asked questions
How many French employees are trained in AI?
According to the KPMG and University of Melbourne study published in 2025, 24% of respondents in France feel trained in AI and 76% have received no training. These are self-assessments, and the French sample size is not published. France ranks 40th out of 47 countries.
Is AI training mandatory for companies?
Article 4 of the EU AI Act, reworded in 2026, asks companies that use AI systems to take measures to foster AI literacy among their staff. It does not impose a specific level per person. Documented training is the simplest way to show those measures.
Does AI stop people from learning their job?
It can, if thinking is delegated along with the task. A 2026 working paper observes that it harms code understanding and debugging among those who rely on it. The answer is not to ban it, but to train people to review and to keep some tasks done by hand while learning.
Where should AI training in a company start?
With an inventory of real uses, including hidden ones, then two or three concrete tasks per role. Good training teaches first what a good result looks like, then the tool, then review and the rules on data.
Sources
- Eurostat, isoc_ai_iaiu (new tab)ec.europa.eu
- INSEE (new tab)insee.fr
- Only 51% have access to the training they need (new tab)prnewswire.com
- Top barrier to integration (new tab)deloitte.com
- Shen and Tamkin (new tab)arxiv.org
- Regulation (EU) 2026/1744 (new tab)eur-lex.europa.eu
Jim · founder of Brain Plus AI
Also in business ai
Stepping back8 min
Does AI make us dumber? What the studies say
Not all the studies doing the rounds on this subject are worth the same. We rank four of them by strength, from a field experiment to a preprint on 54 people, and say what emerges.
Stepping back8 min
Will AI take young people's first jobs?
No disruption is visible across the labour market as a whole, but a signal appears among young people, in the occupations where AI replaces the task. What the studies say, our view, and the real question for an SME: how to train its juniors tomorrow.
Stepping back10 min
The human role alongside AI: knowing how to judge
AI excels at some tasks and is confidently wrong on neighbouring ones. The human's value shifts: knowing where the line runs, keeping your judgement, answering for the result.