An apparent acceleration that masks a deep disconnect
When I analyze recent figures on artificial intelligence in France, an impression of acceleration immediately emerges. Companies are adopting, investments are increasing, discourse is becoming structured. The report “Unlocking France’s AI Potential 2026” announces that 40% of French companies now use AI. At first glance, the signal is positive. Yet, as soon as I compare this data with what I observe in the field, a disconnect appears.
I repeatedly encounter this disconnect in conversations with executives. One of them recently confided: “We use AI every day, but if you ask me what it really changes in how we make decisions, the answer remains unclear.” This sentence alone summarizes the current ambiguity. AI is there, it’s visible, but it’s not yet transformative.
AI adoption: real progress, but still superficial
By cross-referencing data from this report with those from the OECD, the European Commission, and the Stanford AI Index, a more precise reading emerges. Adoption is progressing, but it remains largely superficial. The majority of companies use AI to optimize existing processes, rarely to transform their business model. This distinction is essential. Automating is not transforming.
In interviews conducted for the EntrepreneurIA project, this point comes up repeatedly. An executive from the industrial sector told me: “We’ve gained efficiency, but we haven’t yet changed how we create value.” Here, AI acts as an accelerator, not as a transformation engine. It improves what exists, without necessarily challenging it.
Unequal diffusion: towards a two-speed economy
This situation fuels a form of collective illusion. Adoption indicators give the impression of a homogeneous, almost linear movement. In reality, the diffusion of AI is deeply unequal. Some players move quickly, structure their data, invest heavily, and integrate AI at the heart of their strategy. Others remain in an experimental mode. France Stratégie rightly mentions a risk of economic dualization. In the field, this risk is already perceptible.
Another executive put it bluntly: “AI is creating a gap. Those who truly integrate it are gaining an advantage, the others are accumulating delays.” This polarization is confirmed by the work of the OECD and the Stanford AI Index. Productivity gains linked to AI are not distributed uniformly. They concentrate on a limited number of players capable of scaling up.
The real barrier: organization more than technology
This is precisely where the real challenge lies. The problem is no longer access to technology. The tools exist, platforms are available, models are accessible. The barrier is elsewhere. It lies in organizations’ ability to integrate these technologies into their operations. An executive from the service sector explained it to me this way: “We invested in AI, but we haven’t yet transformed the company so it can benefit from it.”
This gap between technological acquisition and organizational transformation is central. It relates to issues of governance, skills, and culture. The OECD emphasizes that the lack of talent remains a major obstacle, but it’s not just about technical skills. The difficulty also lies in executives’ ability to understand the strategic implications of AI and to lead change.
AI as a revealer of structural weaknesses
This point appears very clearly in the conversations I’ve had. A female executive from the healthcare sector told me: “AI didn’t make us more efficient, it mainly revealed our dysfunctions.” This remark is particularly enlightening. AI acts as a revealer. It amplifies what already works and highlights what doesn’t.
This idea aligns with the Stanford AI Index analyses. The most advanced companies in AI use are also those with strong organizational maturity. Technology doesn’t compensate for structural weaknesses, it makes them visible.
Indicators to be qualified: adoption vs transformation
In this context, adoption indicators must be interpreted cautiously. Saying that a company uses AI doesn’t allow us to understand the intensity of this use or its real impact. I often emphasize this distinction: adoption is a declaration, transformation is proof. This nuance is essential to avoid a biased reading of the situation.
Regulation: a more complex debate than it appears
The regulation question also illustrates the subject’s complexity. The AWS report highlights the weight of regulatory constraints, particularly regarding compliance. This perception is shared by many companies. However, analyses from the European Commission and France Stratégie invite us to nuance this observation. The problem is not only the level of regulation, but its readability and coherence.
An executive from the financial sector recently told me: “It’s not the rule that slows us down, it’s the uncertainty about how to apply it.” This sentence summarizes the issue well. Clear regulation can support innovation. Vague regulation can slow it down.
The decisive challenge: moving from experimentation to scale
Beyond these debates, the central question remains scaling up. Companies experiment, multiply pilot projects, test use cases. But very few manage to industrialize these initiatives. The AWS report estimates that at the current pace, a large portion of companies will only reach an advanced level of AI use in about ten years. This timeframe is out of sync with the speed of technological evolution.
This observation leads me to formulate a simple idea: the gap doesn’t come from excessive AI acceleration, but from organizational inertia. Technologies evolve rapidly, internal structures much more slowly. This asymmetry creates growing tension.
Conclusion: moving beyond statistical illusion
In the interviews conducted for EntrepreneurIA, this tension is omnipresent. One executive summarized the situation this way: “AI is not a technological issue, it’s a strategic issue. As long as we treat it as a tool, we’ll remain on the surface.” This sentence illuminates the nature of the current challenge.
Cross-analysis of different sources therefore allows us to move beyond a simplified reading. AI is progressing, but it’s not yet transforming the entire economy. It creates value, but in a concentrated manner. It opens up prospects, but it also reveals vulnerabilities.
The question is no longer whether companies use AI. It’s about understanding what they actually do with it. Behind adoption figures, a much deeper transformation is taking place, affecting business models, organizations, and decision-making processes.
In your organization, is AI a tool that’s added, or a lever that transforms?
