AI literacy as a meta-skill: a four-domain model for academic management innovation in higher education

As artificial intelligence (AI) becomes embedded in higher education operations, AI literacy is increasingly positioned as a meta-skill enabling institutional innovation; however, its contribution to academic management innovation remains underexamined in China’s private higher education sector. This study surveyed faculty and administrative staff from private institutions in Sichuan Province using a validated four-domain AI literacy (AILit) model—Engaging, Creating, Designing, and Managing—and tested its measurement and structural properties. Confirmatory factor analysis supported strong construct validity and reliability (Cronbach’s α = 0.86–0.93). Structural equation modeling indicated that all four AILit domains significantly predicted innovation outcomes (p < 0.001), with Managing AI showing the largest effect. The model demonstrated excellent global fit (CFI > 0.95, TLI > 0.94, RMSEA < 0.05) and measurement invariance across academic versus administrative roles. The findings suggest AI literacy functions as a strategic, transferable capability extending beyond technical use to include governance, ethical oversight, and institutional alignment, underscoring the need for AI governance training and ethics-based implementation mechanisms. Limitations include the cross-sectional design, self-reported measures, and geographically bounded sampling; future work should use longitudinal, multi-source designs to strengthen causal inference and generalizability.