Each school year begins with a strategic effort to assess the level of competence of students, an effort during which any failure to achieve sufficient acquired levels is considered intolerable. In a context where the heterogeneity of classes represents a major challenge, the integration of artificial intelligence (AI) appears as an opportunity to rethink teaching methods and build a truly "AI-ready" school. This study thus examines how AI tools can help improve mathematical skills and personalize learning among first-year college students.

This design-based study was conducted with a group of students chosen according to a reasoned sample. The goal was to be able to design, test and refine an educational device integrating the learning platform Kwyk.fr. From a diagnostic test, this platform detects the achievements and points to work on for the student and generates an adaptive exercise path.

The analysis of the collected data reveals a notable progression in students' performance, both in terms of consolidation of mathematical skills and in terms of involvement in the proposed activities. These results highlight the potential of AI to support teachers and students in an educational transformation based on real needs, by promoting more targeted learning paths, more effective and better adapted to the diversity of profiles.