University dropout is a persistent challenge that affects students, institutions and the education system in general. To address it, it is necessary to understand the causes and how they are integrated into Artificial Intelligence (AI) models. The proposal goes beyond predicting dropout, it is now required to complement it with a global and local perspective of the model using AI explainability techniques (XAI). Model transparency is crucial to support informed decision making. This study describes an approach to interpret AI models in a Latin American university, using techniques such as Shapley Additive exPlanations (SHAP), and to serve as a tool to design and deploy local and global intervention actions.