As artificial intelligence reshapes business models and the way people work, France is falling significantly behind in its adoption—especially among very small businesses (TPE) and SMEs. A KPMG study highlights a lack of awareness, training, and trust. So how can this gap be closed? How do we democratize AI across companies? Here are our 6 key points.
To explore these questions, we attended the AIM conference in Marseille, where a panel gathered Jean Cattan, Head of “Cafés IA,” Stéphane Hadinger, Director and Head of Technology at Amazon Web Services (AWS), and Véronique Torner, President of Professional Federation Numeum. According to these experts, while France’s delay stems from insufficient awareness, training, and trust, it can still be overcome through a practical, collective, and usage-oriented approach.
A French Delay Rooted in Low Trust and Insufficient Training
European studies show that France performs poorly in AI adoption. According to the latest KPMG report, 76% of employees have received no AI training, placing the country 40th out of 47 worldwide. French companies also rank 25th out of 27 in Europe for digital technology adoption.
Yet a paradox remains: Paris is one of the world’s leading innovation hubs, and French startups show an AI adoption rate above 68%, compared to only 30% among established French companies (42% in the rest of Europe). Infrastructure and talent exist, but usage diffusion remains limited.
For Véronique Torner—citing the Draghi report on EU competitiveness—the issue is not only technological but economic:
“There is a clear link between the digitalization of the economy and its competitiveness. And within that digitalization, AI adoption is a major issue.”
As AI becomes one of the defining challenges of the decade, failing to democratize it would mean falling further behind. Democratizing AI is therefore a strategic priority for competitiveness, innovation, and social cohesion. So, how do we catch up?
Find out more with our infographics on AI adoption in French companies
1. Create Spaces for Dialogue to Defuse Fear
For Jean Cattan, democratizing AI begins with discussion. Through “Cafés IA,” he brings together small groups of 6 to 20 people in an informal setting to understand the technologies, ask questions, express concerns, and co-construct practical uses. The goal is to rebuild trust around a technology that can feel abstract or frightening:
“These technologies, which can generate anxiety, must become objects of dialogue, proximity, and local democratic debate. ”
And to mention the site ‘One AI Each Day’ (Une IA par jour) as a way to discover, explore, and try to understand.
This more horizontal approach complements traditional top-down training, enabling peer-to-peer learning and smoother adoption.
“Education can come from your colleague, from the person you work with every day.”
Cattan recommends that companies dedicate half-days each month to training and collective discussions about AI.
As Véronique Torner adds:
“Surveys show that once employees use AI, they understand its benefits—its complementarity, its ability to make them more efficient. The fear comes mainly from those who have never used it.”
Hence the urgency to raise awareness and train them.
Read also
2. Train at Scale to Unlock Use Cases
For Stéphane Hadinger (AWS), training is the first practical lever for democratizing AI. AWS has trained 200,000 people in France since 2017, with a target of 600,000 by 2030.
Training allows employees to directly see AI’s positive impact on their work. Once they have this initial exposure, fears decrease and companies can begin envisioning improvement and innovation projects.
The shift is already visible. In 2023: companies mostly deployed chatbots. In 2024: they moved toward retrieval-augmented generation (RAG) for documentation search or maintenance.
“From 2025 onward, the focus will be on how AI can drive innovation, reshape business processes, and enable capabilities that were previously impossible.”
One example: the neobank Qonto, which now offers instant working-capital loans triggered directly from supplier payments.
“In banking, issuing a loan usually requires about twenty steps. AI makes it possible to propose it instantly, right when a professional pays a supplier.”
Use cases already exist—what matters now is scaling them.
3. Prioritize Use Cases: Optimization vs. Innovation
AI typically serves two strategic objectives. It’s up to you to make your choice.
Continuous process improvement: Automation of repetitive tasks, workflow optimization, time savings, and reliability. Here, companies should deploy chatbots, RAG systems, and workflow automation.
Innovation and new services: AI enables products and services that were previously impossible—like Qonto’s instant credit.
Optimization projects are easy to evaluate through ROI, but innovation is harder to measure upfront. For Hadinger, the rapid evolution of models requires short experimentation cycles—test and learn.
“Separate these two types of projects and apply the ‘fail fast’ principle: test quickly, adjust quickly.”
Put something in users’ hands, observe reactions, refine accordingly. This helps employees see AI’s tangible benefits early.
Read also
4. AI Agents: Prepare Employees and Find the Right Balance
Habib Messaoudi, Country Practice Leader for Applications, Data & AI at Kyndryl, focuses on preparing employees for AI agents—tools with high engagement potential, autonomy, and decision-making power.
For him, technological deployment alone is never enough. The first step is governance, defining the scope of agents from ethical, responsibility, confidentiality, and human-control perspectives.
“What are we trying to accomplish? What use case, what value, what gains? Only then should we address the technological aspects.”
For him, competencies are the main challenge:
“87% of people recognize that a major shift is coming. The future belongs to those who can think critically with models, execute with models, and decide with models.”
Scaling is the next hurdle. Success depends on finding the right balance:
“Don’t aim for technological perfection—find the equilibrium between innovation and control. It varies from one organization to another.”
Read also
5. Support Through Ecosystems, Partners, and Industry Networks
Even with clear use cases, AI’s technical complexity requires structured support. For Hadinger, the priority is simplifying access so TPEs/SMEs can move from pilots to production.
AWS now offers 249 models from 34 providers, supported by 300 partners in France (20 specialized in generative AI).
“We’re a bit like the e-commerce platform of LLMs.”
This helps companies select the right models, industrialize projects, and secure their data.
For Véronique Torner, support is essential because this “gas race” is like a marathon run at sprint speed:
“Technologies move so fast that a project started today may be obsolete in six months.”
Many companies wonder whether they should wait until technologies mature. Her answer:
“You must start now. If you miss the first train, you won’t catch the next ones.”
Territorial initiatives like the Tour de France de l’IA show how local networks are key to helping companies adopt AI in line with their realities.
6. Clarify Regulation to Build Trust
Finally, there is one last element to take into account to improve democratization of AI in companies: regulation. Does Europe regulate too much? Is regulation slowing adoption? Or can it help democratize AI? For Véronique Torner, a clear regulatory framework is essential:
“French and European businesses support regulation. It gives a framework that protects our values and creates a single European market instead of 27 fragmented ones.”
But excessive or complex regulation can be counterproductive. Hadinger notes that the AI Act includes 113 articles, which is overwhelming for small businesses.
“Large companies can afford compliance teams. SMEs can’t. In France, for every €100 invested in generative AI, €42 goes to regulatory compliance.”
Simplifying and operationalizing rules is therefore crucial. Leaders should integrate compliance from the start and rely on vendors providing ready-to-use tools.
“Most SMEs don’t ask for generative AI itself—they want business solutions that already include it.”
AWS works with software publishers, SaaS providers, and solution vendors to integrate all these technologies so that small businesses can benefit directly without having to invest themselves in regulatory analysis.
Read also
