In France, 54% of companies are struggling to make AI profitable. The fault lies in unsuitable metrics, disconnected from business issues.
Artificial intelligence has moved beyond promise and into execution. In France, as in the rest of Europe, companies are increasing their announcements and experiments. However, a certain disenchantment is beginning to be felt. Despite growing investments, many leaders still struggle to identify tangible economic returns.
This observation is particularly marked in France. According to a recent study conducted by the Ponemon Institute1 on behalf of OpenText, including French organizations among its respondents, less than half of French companies (46%) say they are confident in their ability to demonstrate the return on investment of their AI initiatives – compared to 61% in North America. These results highlight that realizing the full potential of AI requires strong governance, clear measurement frameworks, large-scale execution and secure, well-regulated information management.
The gap between ambition and operational reality
The contrast is striking. Executives show strong confidence in their ability to leverage AI, while IT and data teams paint a more nuanced picture. According to a recent PwC study, 67% of executives believe their information systems are ready for AI — an opinion rarely shared by CIOs, who point to well-known obstacles: insufficient data quality, accumulated technical debt and complex application environments.
This discrepancy fuels persistent incomprehension. AI is often perceived as widely deployed, when in reality, less than a quarter of French companies have industrialized it on a large scale. Only 21% use it in a structured way in critical functions like sales, marketing or customer service. Initial enthusiasm gradually gives way to a realization: the real bottleneck is no longer innovation, but execution.
The illusion of technical indicators
In this context, the way AI performance is measured becomes crucial. Too often, organizations rely on technical or activity-related indicators: number of use cases, volume of automation, intensity of tool use. These metrics (such as the percentage of code generated by AI) may be reassuring, but they do not answer the central question that executive committees are asking: how does AI concretely improve the company’s performance?
Worse still, these indicators can mask counterproductive effects. Rapid but poorly managed adoption can increase IT complexity, deepen technical debt or weaken application security. Teams may appear more productive, without there being any real improvement in service quality, operational reliability or customer satisfaction.
What leaders really want from AI
For general management, AI is not an end in itself. It must serve very concrete priorities: accelerate time-to-market, secure operations, strengthen resilience to incidents and manage the increasing complexity of information and regulations.
In other words, AI is judged by its ability to improve the overall functioning of the company. It must contribute to more reliable systems, shorter decision cycles and better business continuity. This is where its economic credibility comes into play.
Rethink performance indicators
To get out of the current impasse, companies must rethink their indicators and adopt a framework more closely aligned with business issues. Three dimensions stand out.
The first dimension is the ability to deliver faster without weakening existing systems. Shorter and more frequent deployment cycles reflect an organization capable of using AI to streamline processes while maintaining high quality standards.
The second is operational robustness. Fewer incidents, faster service recovery and fewer patches in production are clear signals that AI is helping to stabilize digital environments rather than making them more complex to manage.
Finally, organizational maturity. The Ponemon–OpenText study shows that most organizations (54% in France) experience difficulty supporting innovation and business transformation because IT and business objectives are not aligned. Without this alignment, as well as strong data quality and effective AI governance, AI remains confined to promising but difficult to industrialize pilots.
From media hype to economic proof
The current phase marks a turning point. The enthusiasm for AI is fading, giving way to a stronger demand for results. Leaders now expect proof, not technological demonstrations. To meet this expectation, companies must learn to translate AI contributions into clear, comparable and actionable performance indicators.
Those who succeed in this change of perspective will not just experiment with AI. They will be able to demonstrate, with supporting data, that this technology really contributes to growth, cost control and business resilience. In a tense economic context, this ability to objectively measure value will make all the difference.
