70% of data projects fail. Not for lack of training, but because the distance between data and decision is structural. Native AI profiles eliminate it.
No, change management will not be enough to move business teams from data to action. For more than 20 years, organizations have been building their indicators, installing business intelligence tools, then setting up training programs to have them adopted by their business teams. In the majority of cases, the result is always the same: a partial and episodic appropriation of the tools, and a succession of iterations with the data teams to converge on the real need which, despite their commitment, end up saturating them. It is time to understand artificial intelligence not as a technical agent added on top, but as a native profile of the data platform, anchored in the processes and uses of the user’s function. This is the key to data adoption.
A problem as old as traditional BI
It’s a pattern that all CIOs, data analysts and business directors know well: the data team builds a warehouse, models the indicators, configures the dashboards, and ends up giving the keys to the tool to the Business Lines. And this is where everything gets complicated. A gulf is created between data production and operational use.
Companies then set up a change management program to support them in the appropriation of these new tools and indicators. But the more time passes, the more the gap widens between data and business use. After 12 to 18 months of training, less than 30% of trained teams are autonomous and 70% of data projects fail to generate lasting business value. This blockage is not educational: it is structural.
The adoption wall
Change management often starts from an implicit premise: if teams do not adopt the data, it is because they are missing something: training, motivation, an additional explanation. Acting on the right lever would be enough to promote data appropriation. This reasoning is not wrong. It is simply missing the real problem.
A sales director who no longer opens his dashboard or who does not explore his indicators has not given up on managing his activity through data. He waived whatever it cost him. Between the question he asks himself one morning (why my sales to population His only recourse remains to contact a data expert, again and again. The bottleneck is not going away. He moves.
Native AI profiles as in-use infrastructure
Change management can treat the symptom of resistance, but not resolve the cause: the distance between data and decision. The paradigm shift is radical: moving from observation to business use is not a matter of humans, but of the data & AI platform.
An AI profile, unlike a general AI agent, is an intelligence anchored in the processes and language of a specific function (sales, finance, HR). More than responding to a request, he understands the context in which it emerges.
Giving back control to the professions requires modernizing the data environment with a unified platform that covers the entire data cycle, and within which AI profiles are natively integrated at each stage (and not on top of it) to assist the user throughout their exploration and analysis work. They do not simply guide him in the path between the data and the decision: they eliminate a large part of it.
Present in the right place, at the right time, and exactly adapted to business use
Designed as an extension of the profession, these AI profiles support the user continuously, at the right time and in the right context. Designed to adapt precisely to the role of each user (commercial, CFO or other), they support them in their real use cases, as they arise.
They dialogue with him, refine his intentions, clarify his choices and instantly carry out explorations and analyses, to the point of testing his intuitions and allowing him to take action. All this within the company’s data platform, in on-premise mode, without any data leaving its sovereign scope.
By building on the foundations of data teams, they absorb technical complexity and streamline the entire journey. Result: we move from slow and iterative exploration to fluid and continuous management, where the user progresses directly from the question to the decision.
The real challenge is not only to produce indicators, but to enable an autonomous and immediate transition from observation to action. By making the exploitation of real-time data accessible in natural language, the AI profiles remove the technical barrier, while the data platform guarantees the reliability of the executions to avoid any hallucinations.
Break in posture and sovereignty
This approach constitutes a break in posture more than in technology. Because each AI profile is dedicated to a specific use, anchored in the processes of a function and present at each stage of the data cycle, from production to exploration, from catalog to decision, the user no longer needs to be trained: he builds his own understanding, through his questions, his explorations, his arbitrations, alongside the right AI profile.
A vision that is all the more different because the question of sovereignty is central to public organizations and large French and European groups. Deploying native AI profiles, in on-premise mode, within a sovereign data platform, without transit to third-party infrastructures or dependence on American hyperscalers, offers an alternative that the dominant players in the global market cannot replicate. By unifying governance, exploration and action, we no longer train users in data: we finally make data usable.
