Agentic AI Divide in Higher Education

Abstract

Higher education institutions are expanding their use of agentic artificial intelligence (AI), and fundamental questions about equity, privacy, and fairness have emerged. Unlike earlier rule-based systems, agentic AI autonomously manages complex tasks including adaptive instruction, academic risk prediction, and institutional decision-making. This systematic literature review synthesises findings from 28 peer-reviewed studies (2021–2026) to examine how agentic AI creates or deepens inequalities among marginalised student populations. Three intersecting challenges are identified: (i) an AI literacy divide that prevents equitable engagement even when access exists; (ii) data governance deficits that erode student trust through opaque surveillance practices; and (iii) algorithmic bias that can silently reproduce structural inequities. Results confirm that access alone is insufficient; meaningful and fair use requires targeted AI literacy training, student-centred data governance frameworks, and bias-mitigation strategies embedded in institutional policy. This review calls for governance models that position students as rights-bearing participants and for longitudinal research extending beyond high-resource institutional contexts.