Analytical Study of AI-Enabled Data Platform Integration in Higher-Education Institutions for Improved Student Success and Research Performance Outcomes

Abstract

Higher education increasingly demands timely, reliable analytics. At many universities, separate data systems for learning, research, finance, student, and identity services create significant integration, semantic, and auditability challenges. This paper contrasts the use of a governed, AI-powered data platform to improve decision support for student success and research performance. Researchers integrated 32,593 students, a stratified panel of 300 four-year U.S. universities from 749-771, 18 purposively selected stakeholders, and 200 standardized prompts for evaluating LLMs into a Design-Science Research and mixed-methods approach. The platform brought together cloud ingestion, lakehouse storage, master data management and semantic KPI governance, knowledge graphing, predictive models, and retrieval-grounded LLM services. Data completeness improved from 91.2% to 98.4%, KPI agreement improved from 72.6% to 96.8%, and reporting latency improved from 26.4 hours to 2.1 hours. Student-risk AUC increased by 0.09 to 0.83, research-forecast MAPE decreased from 19.6% to 11.8%, and time to insight decreased by 64.3% with the assistance of an LLM. Lineage completeness was 97.0%, and cross-domain reconciliation was 98.2%. These results show operational improvements but do not provide concrete causal evidence of improvements in retention or research funding. The results suggest that co-developing integration, semantics, governance, and human oversight is essential for successful implementation of institutional AI. Further study is needed to validate causal, longitudinal, and cross-institutional effects in a production-scale deployment