Sanofi — Making generative AI useful and trustworthy in healthcare

MedIS, Sanofi’s Medical Insights Engine, uses GenAI to produce insights that help detect care gaps (“GenAI insights to detect care gaps”).

The real challenge is human. An expert does not use a recommendation they cannot explain. In a regulated environment, an AI perceived as a black box stays on the shelf, however well it performs. The question to solve was therefore one of adoption: how do you get demanding professionals to understand what the AI is proposing, question it, and then make it their own?

• Start from the field. The starting point is users’ real decisions and their day-to-day work. Making GenAI insights trustworthy for healthcare professionals begins with a close understanding of the reality of their work.
• Co-create with the experts. Co-design workshops with business and scientific experts anchored every experience choice in a real decision need.
• Address the right friction: uncertainty. The main barrier was doubt. Each iteration aimed to close the gap between what the AI claims and what the expert can verify for themselves: making every insight readable and verifiable.
• Keep the human in the loop. The AI proposes, the expert judges. The decision stays with the expert, and this principle guided every trade-off.
• Iterate at the pace of trust. Exploratory concepts, user feedback, adjustments: the design progresses as trust is built.