Behind on AI? What the 2026 figures say

“We are behind on AI.” The sentence comes up in almost every executive committee. It assumes that the others have adopted AI and you have not. The figures published in 2026 say something else.

What the Banque de France and Insee measure #

Two public surveys came out six weeks apart. They do not measure the same thing.

The Banque de France adds two details. The gap by size is small: 64% between 20 and 49 employees, 83% above 1,000. And this use remains mostly limited or experimental.

Insee, for its part, measures growth: 6% in 2023, 18% in 2025, three times more in two years. The institute also notes that French companies have caught up with European companies of the same size, except those with fewer than fifty employees, give or take two points.

These two measures neither add up nor contradict each other. They have neither the same scope, nor the same year, nor the same definition. Read together, they say one simple thing: adoption is no longer the issue. The Banque de France concludes as much itself, writing that the challenge now lies “less in the adoption of AI than in its capacity to transform” the organization (our translation).

Your employees have moved ahead #

Look at the people. According to the Digital Barometer, carried out by Crédoc and published in February 2026, 48% of the French population uses generative AI. Among managers and professionals, it is 76%. The study was conducted in June 2025.

The Banque de France sees the same mechanism from the company side: generative AI can be used directly by employees, through standardized tools. Your employees did not wait for your roadmap.

Why “behind” is the wrong word #

The word “behind” pushes you to buy more and launch one more project. Yet what these figures describe is usage that has moved ahead of the organization.

AI is already in your company, probably in more software than you could name from memory. It came in through your vendors, who add features to the suites you already pay for. Through your administrators, who opened a pilot or a trial. Through your employees, who use the tools they know.

You do not have an adoption problem. You have a knowledge problem: AI is there, and nobody keeps the list.

The three questions that tell you where you stand #

Ask them at your executive committee.

1. Which AI features are active in our company, in which software, and since when?
2. How much does AI cost us per year, across all vendors, and what share of what we pay is really used?
3. Who is allowed to use what, and with which data?

If the answer is “someone would have to go around the departments”, you are not behind. You do not know where you stand. It is not the same situation, and it is not handled the same way.

What changes when you know #

Falling behind is fixed by spending. A blur is cleared by counting: one line per AI, with its vendor, the door through which it came in, its date, its cost and its status.

With this list, you answer the three questions in a few minutes. You decide line by line what you keep and what you turn off. And when a vendor announces a change, you know right away whether it concerns you.

To draw up this list, see the AI inventory in a company and its three doors.