Master AI to be employable, offer AI to be attractive

First, the obvious: master AI to stay employable

Twenty years ago, everyone wrote "proficient in Microsoft Office" at the bottom of their CV. AI is going the same way, only much faster. According to PwC's global barometer published in June 2026, which analysed more than one billion job ads, postings that ask for AI skills are growing eight times faster than the market as a whole (69% against 9%).

At Mister IA, we see it with our clients. Companies that have trained hundreds of employees do not want to start from scratch with every new hire: they want the salesperson or the lawyer they recruit to already know how to use the tools on day one.

Employees have understood this too. In 2023, 150 people bought an AI course with their CPF, France's personal training account, over the whole year. In September 2026, around 2,500 did so in a single month.

Caisse des dépôts, MonCompteFormation open data, analysis by Mister IA · 2023 to July 2026; August and September 2026 estimated by Mister IA

That is still less than 2% of all CPF enrolments, and there is still a long way to go. But the direction is clear.

Next, what fewer people see: offer AI to be attractive

Let us look at the problem from the other end. More and more candidates arrive already used to working with AI, a habit picked up in their previous job or during their studies: six in ten students and recent graduates use it every day, according to the Ipsos bva barometer for SKEMA and EY from February 2026. These candidates will look at what the company puts in their hands, and two things will matter.

Equipment first. Is there a tool at all? Only 45% of private-sector employees say their company has officially rolled one out, according to the Ipsos bva observatory for Greenworking published in September. Which tool? An assistant built in-house is no match for the paid versions of Claude, ChatGPT or Copilot. How much can you use it? Someone used to working without counting will not put up with a quota that runs out mid-afternoon. And finally, what is it connected to? A salesperson who used to ask their AI to draft a proposal straight from the CRM will not accept copying client records by hand. The same goes for the management controller whose AI read the ERP directly, or the accountant whose AI was plugged into the accounting software.

Culture next. Does the company encourage the use of AI, or merely tolerate it? Does it train its teams to use it? In practice, culture is the manager who uses it in front of their team, the examples colleagues share with one another, and the time people are given to learn.

Put the two together and you get four types of company.

Mister IA matrix · AI culture × AI equipment, four company profiles

With good tools but no support, you attract people who are already self-sufficient, without managing to keep them. With the culture but without the tools, you do not even get candidates through the door, and the people already there end up frustrated. Only companies that have both attract and retain talent.

It is reminiscent of remote work, which has become a deciding factor for many candidates. According to Apec, eight in ten managers who work remotely would be unhappy if it were taken away, and half of them would look elsewhere. My bet is that AI will follow the same path, and that job ads will soon state the AI tool provided, just as they now list the number of remote-working days.

A challenge, especially for large groups

This challenge will face every company, but large groups first, because they are falling behind on both equipment and culture. The statistics seem to say the opposite: according to Insee, France's national statistics office (July 2026), 58% of companies with 250 or more employees use AI, against 15% of those with 10 to 49 employees. But that figure counts companies that have at least one tool, not employees who have a good one.

What we see at Mister IA is quite different. First, the share of employees with a paid licence is lower in large groups than elsewhere, partly because of their size and the complexity of their IT systems. Second, when they are equipped, it is often with an underperforming in-house assistant, or with the free version of Copilot, which cannot compete with the paid versions of Claude or ChatGPT. Finally, the culture lags behind: use is often regulated before it is encouraged, and gaps sometimes widen from one team to the next.

An account manager who has spent two years at a 500-person mid-sized company, with Claude Team connected to the CRM, will not join a large bank to find a capped tool cut off from everything. Or they will join, and quietly keep using their personal subscription: only 37% of users stick to the tools approved by their employer, according to Ipsos bva for Greenworking. By locking everything down to protect its data, a company ends up exposing it more.

CIOs have good reasons to be cautious: a bank or a hospital does not hand its data to just anyone. But that caution has a cost, and HR will pay it, in roles that stay unfilled. The tug of war between HR and IT has begun, and HR directors have every reason to lead it.

Vincent Pavanello, co-founder of Mister IA

← All analysis