Does artificial intelligence have the time of democracy? Difficult to make room for this human, slow and groping process, within a technological revolution as dazzling as it is powerful. How to take the time for deliberation when, in a few days, the Chinese Deepseek application, as efficient as Chatgpt for a derisory cost, causes a stock market storm and shakes an American hegemony that we thought was acquired? And yet, for both ethical and practical reasons, a place must be reserved for democracy in the deployment of AI.
Let us hear: for a company, the question of the opportunity does not arise. AI is a turning point that is essential. Beyond the expected efficiency and comfort gains, AI, in particular generative, establishes a new experiential standard that it is impossible to ignore. The question of how remains entirely. On the one hand, technologies are not mature and the state of the art is constantly evolving, the Deepseek twist illustrates perfectly. Poor choice of investments can lead to gigantic waste of resources. On the other hand, the success of the deployment of AI will depend for many on the capacity of the social body of the company to grasp it. Resistance, disinterest or muddled frenzy: How to avoid counterproductive reactions and effectively carry out this global transformation?
Put up safeguards
This is where democracy comes in. Without slowing down technical experiments, taking the time of strategic deliberation with employees and stakeholders of the company makes it possible to gain alignment and, therefore, in performance. It is a way of sharing knowledge, approaching fears, setting up safeguards and defining common objectives.
This is the choice made by MAIF by organizing the first “salaried agreement” dedicated to AI. Like citizen conventions for climate or end of life, we have drawn lots of employees who, for four days, were trained by experts before writing proposals for the company. There have been around forty measures, both ethical and operational, and a dialogue has engaged with the general management, the board of directors and the unions to integrate it into the company’s strategy.
What were the conditions for the success of this experience? First, great freedom. No question was taboo and the conventionals had no constraints in drafting their proposals which could both relate to employment, energy sobriety as well as identification of use cases.
Beware of untenable promises
Above all, a clear commitment was made by general management. Lighted by the experience of the Citizen Convention for the Climate, at the origin of a lot of hope and then a deep disappointment, we never suggested that the Convention was going to determine the company’s strategy in matters of ‘Ia. The Convention makes an important contribution with a concrete translation, but it is the direction which ultimately prioritizes, contrasts and fixes the course. Getting involved in a resumption of all “filter” proposals would have been an untenable promise and would have fueled a confusion of roles that does not contribute to a serene democratic practice.
What were the benefits for the company? First, a gain of trust, then an acceleration. Confidence arising from the reassuring and transparent responses provided by the general management on employment, ethics, working conditions and the place of humans in our profession. Acceleration arising from collective enthusiasm and ideas from the field to meet the needs of employees and customers.
The revolution of generative artificial intelligence is certainly spectacular, but its deployment will extend over at least a decade. To go on good rails, taking the time to discuss it collectively is not so luxurious. This is partly the meaning of the World Summit on AI to be held on February 10 and 11 in Paris: innovation and deliberation feed each other.
* Pascal Demurger is managing director of the Maif group and co -president of the Impact France movement.
.