Generative artificial intelligence is now invited into decisions which, so far, have been exclusively human discernment. No longer confined to assistance tasks, it is starting to weigh on structuring choices for businesses. This delegation, still marginal in France, challenges as much as it intrigues. Some organizations have nevertheless taken the plunge and experimenting, in real conditions, what it means to entrust to an AI to decide.
A real delegation, not a simple gadget
When Charles Guilhamon, founder of Faguo, decides to reposition a line of products for the start of the 2024 school year, he is not only based on his internal teams. It also submits all sales data, customer feedback, research trends and competitive benchmark to an internal generative AI. This AI does not provide him with a summary report, but an argued recommendation, accompanied by encrypted projections, in the segment to be prioritized. After validation, the brand follows this orientation: abandonment of a loss range, reinforced highlighting of recycled products on the e-commerce canal.
For the management team, it is not a test, but a real temporary transfer of the decision. “The choice was ready to be decided, but we wanted to see if the AI would offer a founded alternative. And it posed a hypothesis that no one in the team had considered, with a very structured logic,” says Guilhamon. The result has proven conclusive, sales of the new range exceeding the objectives at three months.
Optimizing the time of decision -makers
The Lyon SME Yelda, specializing in vocal assistants for companies, joined a decision -making AI in 2023 in the management of its commercial offers. Faced with a dilemma: maintaining a historic formula or launching a dynamic subscription pricing, the marketing team has chosen to leave the Sltt, from cross simulations over six months. AI, nourished by internal and external data, has proposed an unprecedented hybrid model. Again, it was an operational decision that resulted from it, not a simple advisory opinion.
According to their director general, this delegation made it possible to unclog internal committees often paralyzed by the search for consensus. “What we test is not only the effectiveness of the tool, but also a new mode of governance: faster, more fluid, based on iteration rather than waiting for perfect alignment.” Piloting by AI is not replaced by strategic vision, but it releases brain time for long -term arbitrations.
A decision support tool that has become co -piloting
At Klaxoon, the collaborative platform based in Rennes, generative AI is no longer confined to technical assistance roles. It intervenes in the very design of offers and commercial action plans. During an overhaul of their customer onboarding tunnel, an AI generated several user routes according to the typologies of accounts. The team retained one of the scenarios developed by the tool, with minor adjustments. Result: a conversion rate doubled in three months.
The company’s product director is now talking about “strategic co -piloting”: a system where AI offers, argues, projects. The human valid, adjusts or rejects. But the decision -making dynamics are well initiated by the machine. This model today inspires other French scale-ups, especially in healthtech and edtech, who seek to combine their business vision with the ability to compromise artificial intelligence.
Limits still clear but scalable
Not all leaders are ready to give up from power to an algorithmic entity, even carefully trained. At Manomano, the founders tested an AI system to optimize logistics on certain critical flows. But when the system recommended concentrating stocks in a specific region, the team preferred to delay. “The analysis was mathematically solid, but it did not take into account a particular local context that only a human reading could apprehend,” explains a framework of the operational management.
This is one of the main reservations expressed by French leaders: AI reason in logic, sometimes with more rigor than humans, but it remains blind to weak signals which are the ground or entrepreneurial intuition. However, these same leaders recognize that the tool gradually refines its proposals as it is faced with reality, corrects its biases and learns refusals.
Towards a new risk culture
Delegating a decision to an AI is not denying the manager’s responsibility is to assume another way to exercise it. At Alan, management has experienced a partial reorganization of sales teams on the basis of suggestions generated by IA: redefinition of sectors, reallowing of leads, new scoring criteria. A significant part of the proposed adjustments has been applied. In the following months, commercial performance increased by 17 %. Management sees it as a clear indicator: it is not a question of having confidence blindly, but of recognizing that certain technical decisions can be better taken by a system which has neither ego, nor fatigue, nor emotional biases.
This type of test can only succeed if the company agrees not to always be right, right away. The relationship to error changes: it becomes data-driver, iterative. The decision is no longer imposed at a time, it is built by successive adjustments. For the leaders who engage in it, it is not only a time saving, it is a deep transformation of their posture. No longer that of the omniscient decision maker, but the conductor who chooses the right instruments – and sometimes accepts to let AI play the first note.
An organizational learning in real time
One of the most interesting benefits of these experiments lies in the way in which the teams themselves gradually appropriate the recommendations generated by the AI. At OpenClassrooms, an internal unit dedicated to educational analysis began to integrate the suggestions of an AI to adapt training formats according to the behavior of learners. Originally considered as a complementary tool, this AI is now perceived as a real catalyst for collective decisions. The teams are no longer content to validate or not to validate its proposals: they use it as a starting point to restructure working methods, rethink courses or initiate transverse sites. It is this dynamic of continuous organizational learning-triggered by confrontation with another logic-which perhaps marks the deepest change in managerial culture. AI does not replace individuals, it changes their way of thinking together.
