How do you integrate tools that know nothing about your organization, yet are already changing the way people work there? This is the question explored in Gérer les IA génératives. Cadres, méthodes et dispositifs (Managing Generative AI: Frameworks, Methods and Tools; Presses des Mines, 2026). The book stems from the work of the Bauhaus des Transitions, a Mines Paris-PSL Chair that studies the management of the unknown. It is intended for managers, executives, scientists and civil servants facing contemporary challenges. The book advocates an approach known as design-oriented management (gestion conceptive), which consists in collectively designing what does not yet exist rather than optimizing what is already known. Antoine Bordas, co-editor of the book and postdoctoral researcher at i3-CGS (Mines Paris – PSL), discusses the approach.
Generative AI tools are versatile and powerful, but they enter organizations knowing nothing of their professions, processes or users. This genericity is their strength, but it also leaves them without bearings. For them to create real value, they must be continuously anchored in a context. Their role is also changing: they no longer simply carry out tasks but take part in writing, decision-making and innovation. This raises new questions. How should we work with these new tools? What impact will they have on how companies are organized? Who bears responsibility? Current managerial, organizational and governance frameworks are still struggling to provide answers.
To make headway on these questions, the team chose to go into the field. “Since 2024, we have been studying how to manage generative AI effectively through workshops. We don’t just do academic research there; we immerse ourselves in real-world situations, alongside practitioners,” explains Antoine Bordas.
First, some clarification is needed. “The term ‘generative’ covers very different tools, from image generators to document synthesis engines. A search engine retrieves existing information, whereas a generative AI infers its answer. It interprets the request, draws on latent knowledge, then shapes the output,” he explains.
“Generative AI does not come up with radically new solutions. It does, however, excel at combining known solutions and identifying risks. Understanding it also means understanding what it does to the organization that adopts it, and how that organization must evolve to accommodate it,” he continues. Determining what can be expected of it is therefore essential.
Far from replacing teams, these tools are challenging management practices. This is the focus of the workshops held under the auspices of the Bauhaus des Transitions, which brought together professionals from large companies, start-ups, public institutions and regulatory bodies, as well as researchers. Concepts, frameworks for reasoning and real-world cases served as the basis for this collective work. “We want these workshops to prompt reflection on what good management looks like when generative AI enters the organization,” Antoine Bordas emphasizes.
L’un des cas présentés dans l’ouvrage et abordés en atelier porte sur l’industrie automobile. Les inventeurs interrogés chez l’un des constructeurs indiquent que l’IA est rarement utilisée pour concevoir de nouveaux produits. Elle sert surtout en fin de développement, pour corriger des problèmes repérés lors des tests. Les auteurs de l’enquête parlent d’« exploitation inventive », une démarche créative qui ne remet pas en cause l’existant. Deux types de freins limitent un usage plus exploratoire, à savoir des données abondantes mais difficiles à exploiter et une organisation peu favorable. La responsabilité est partagée avec les équipementiers, aucun processus n’est dédié à ces usages, et les inventeurs se sentent peu légitimes à porter des innovations plus ambitieuses.
The book builds on these discussions, drawing on research findings. What emerges is an idea that challenges the conventional view of management, yet lies at the heart of the concept of design-oriented management developed at i3-CGS. Managers are no longer simply those who decide and control; they become the designers of an environment in which learning and exploration are made possible.
A valuable read for executives and managers who need to integrate these tools now. Meanwhile, the work continues. “A new workshop format and a broader research programme to characterize good management of generative AI are already under way,” concludes Antoine Bordas.
Antoine Bordas, Pascal Le Masson and Benoît Weil (eds.), Gérer les IA génératives. Cadres, méthodes et dispositifs [Managing Generative AI: Frameworks, Methods and Tools], Presses des Mines, 2026.