This research explores how machine learning (ML) techniques could help improve governance in the Moroccan public sector. To this end, we analyzed a dataset comprising audit reports, feedback from citizen surveys and administrative performance indicators. Using Random Forest (RF) and Support Vector Machine (SVM) algorithms, we aimed to identify key links between certain governance characteristics and the actual effectiveness of organizations. In this respect, the analysis suggests that Random Forest models better reflect the complexities specific to the public sector than do SVM models. In particular, it shows that openness and accountability are particularly important levers of performance, mainly, it seems, in fostering public trust and encouraging public participation. Other factors such as management style, innovation efforts, respect for ethical conduct and strategic vision also play a significant role, often in interconnected ways. Given this state of affairs, these findings highlight the value of AI approaches to better understand complex administrative dynamics. They also support the idea that investments in digital transformation and the strengthening of control mechanisms are relevant. Nevertheless, this work opens up prospects for future research, such as exploring more advanced AI models or conducting cross-country comparative analyses, to deepen our understanding of effective public governance strategies, in Morocco and potentially elsewhere.