With the profound application of artificial intelligence in higher education, university evaluation systems are undergoing a significant paradigm shift from a “management-oriented” model toward a “governance-oriented” approach. AI provides new technological foundations for process-based evaluation, value-added assessment, and multi stakeholder collaboration, yet its adoption still faces constraints from institutional, organizational, technical, and value related dimensions. Institutional inertia and unclear data ownership hinder the adaptation of existing evaluation systems to intelligent environments; ambiguous responsibilities and limited trust among stakeholders weaken participation in AI-based evaluation; persistent issues such as data fragmentation, algorithmic bias, and poor model interpretability further heighten uncertainties in evaluation outcomes. Meanwhile, tensions between technological efficiency and the humanistic values of education remain unresolved. In response, this study proposes strategies including constructing a flexible institutional framework, reshaping collaborative governance mechanisms, strengthening agile technical governance system, and fostering a culture of trust, aiming to offer feasible pathways for reforming university evaluation systems. The study argues that only through the coordinated advancement of institutional, technical, and value dimensions can AI truly empower educational evaluation to be more fair and developmental, to contribute to the modernization of higher education governance.