LLM-Augmented Academic Analytics: A Retrieval-Augmented Generation Framework for Real-Time Student Performance Intelligence in Higher Education

Abstract

Higher education institutions face a persistent challenge in translating large volumes of student performance data into timely, actionable intelligence for faculty and academic advisors. Conventional learning analytics dashboards rely on pre-aggregated historical datasets and static visualizations that are incapable of addressing context-specific, real-time queries. This paper proposes the Retrieval-Augmented Academic Analytics Framework (RAAF) — a novel architecture that integrates Retrieval-Augmented Generation (RAG) with existing institutional BI infrastructure to enable natural language querying of live student performance data. Drawing on foundational work in enterprise RAG-BI integration (Pathoori, 2025), this study extends the paradigm to the specific constraints and governance requirements of higher education environments, including FERPA compliance, multi-source academic record systems, and the unique challenge of hallucination risk in high-stakes advising contexts. The RAAF framework is validated through a conceptual case study at a private liberal arts university, demonstrating a projected 64% reduction in time-to-insight for academic advisors and a structured approach to LLM-generated output governance. Findings suggest that RAG-based academic analytics represents a meaningful evolution beyond static dashboards and carries significant implications for early intervention systems, retention analytics, and institutional decision support.