Adopting AI at Work Isn’t Enough to Boost Productivity

France spends almost as much as the United States on IT. French people make wide use of generative AI on an individual basis. French research remains close to the European average. Yet productivity is stagnating, and French companies struggle more than their American counterparts to integrate AI across all their operations.

This is the paradox explored by Antonin Bergeaud and Simon Bunel in the first Policy Paper from LIFT, the Laboratory for Innovation and the Future of Work. Since 1995, hourly productivity has grown by 28% in France, compared with 66% in the United States. Yet in 2025, France is devoting 6.5% of its GDP to IT spending, against 6.8% across the Atlantic. So where does the problem lie?

The data suggest looking beyond access to technology. Adopting AI is one thing. Transforming the company to take full advantage of it is quite another.

At first glance, France is hardly a country resistant to artificial intelligence. In 2025, 37% of French people aged 16 to 74 use generative AI, for personal or professional purposes, compared with an EU average of 33% and 32% in Germany. At work, however, the picture changes: only 28% of French workers use generative AI, compared with 43% in the United States. And those who do use it spend around 7% of their working time on it, against 13% across the Atlantic.

On the business side, Eurostat estimates that 18% of French companies with ten or more employees use at least one AI technology, compared with 26% in Germany. The gap is concentrated in SMEs: among companies with 10 to 49 employees, the adoption rate is 15% in France versus 23% in Germany. For large companies, however, the rates observed by Eurostat are comparable in both countries. But the adoption rate tells only part of the story.

Adopting AI Is Not Yet Integrating It

It is when we look at what companies actually do with AI that a deeper difference emerges. French companies that have adopted generative AI use it as much as their American counterparts for product development. 

But they integrate it far less into functions that cut across the organization:

• internal processes: 56% in France versus 84% in the United States
• marketing and sales: 33% versus 67%
• human resources: 25% versus 48%
• customer relations: 21% versus 47%

The difference is important. Using an AI tool for a specific task does not necessarily require transforming the company. Integrating it into HR, marketing, customer relations or internal processes, on the other hand, may require data to flow across the organization, procedures to change and work to be reorganized. In other words, technological experimentation can be quick. Organizational transformation is much less so.

When Technology Moves Faster Than Organizations

The contrast becomes clear when employees are asked. Nearly seven in ten French workers (69%) are neither encouraged by their employer to use AI nor provided with the tools or training to do so. In the United States, the figure is 44%. Conversely, only 17% of French workers are both encouraged and equipped, compared with 42% in the United States.

The issue therefore goes beyond the mere availability of a tool. To deliver productivity gains, adopting a technology may require complementary investments in data, software, training and organization. And transforming an existing organization is costly.

In France, a restructuring costs the equivalent of 38 months’ salary per affected employee, compared with 31 in Germany and 7 in the United States. Introducing new technologies may also require prior information and consultation of the CSE, the company’s works council. In 2025, two French courts suspended the rollout of AI tools because the consultation process had not been completed, one of which was still in its pilot phase.

Constraints related to data and legal risk may also weigh more heavily on certain functions, particularly human resources and customer relations.

What If the Problem Isn’t a Lack of Skills?

Another intuitive explanation does not hold up well against the data: that French employees simply do not know how to use these technologies. Yet the share of IT specialists in employment is almost identical in France and the European Union: 4.9% versus 5%. And among French people who do not use generative AI, only 13% say they do not know how to use it. Some 62% say they simply have not needed it.

Above all, the relatively high level of individual AI use in France shows that the tool is not unfamiliar to workers. The gap therefore seems to lie more in the conditions that allow companies to move from individual use to use that is integrated into the organization.

Buying Technology Isn’t Enough to Transform a Company

This finding is not entirely new. During the previous wave of information technology diffusion, the European subsidiaries of American multinationals were already managing to extract more productivity from the same technologies than European companies. Research cited by the authors links part of this advantage to their management practices.

Artificial intelligence now seems to be raising a similar question. A technology can be available, used and even widely diffused without immediately transforming productivity. Between the two lies the company itself: its processes, its management, its data, its skills, its rules and its ability to reorganize work.

The renewal of the business landscape could also play a role. A company founded today can design its operations around AI from the outset, whereas an established company has to change existing procedures, information systems and contracts. The authors, however, present this mechanism as a hypothesis that would require additional data to be tested.

From Adoption to Transformation

The Policy Paper thus calls for moving beyond a reading of the AI revolution based solely on the number of users, the number of equipped companies or the amounts invested. The question becomes: what needs to change within the company for AI to truly change the way people work?

The data gathered by Antonin Bergeaud and Simon Bunel suggest that a large part of the answer lies in complementary investments, support for employees and the ability of organizations to adapt their processes. AI can enter everyday use very quickly. Transforming the company to take full advantage of it takes longer.