RETRIEVAL-AUGMENTED GENERATION (RAG) CHATBOT AS AN
INFORMATION MANAGEMENT SUPPORT TOOL: A STUDY WITH ADMINISTRATIVE
STAFF AT THE SERIDÓ HIGHER EDUCATION CENTER (UFRN).
Retrieval-Augmented Generation (RAG); Agentic Artificial
Intelligence; Information Access; Knowledge Management.
Access to information in Higher Education Institutions (HEIs) is frequently hindered by the fragmentation of access channels and information overload. At the Seridó Higher Education Center (CERES/UFRN), this scenario was manifested by the dispersion of procedural and administrative information across multiple repositories, generating inconsistencies, rework, and difficulties for the academic community. Given this context, this research aimed to develop a virtual assistant based on an Agentic Retrieval-Augmented Generation (RAG) architecture to optimize access to institutional documents and their information. Methodologically, the work was grounded in Design Science Research (DSR), structured in four phases: (1) needs diagnosis with administrative staff; (2) curation and structuring of the documentary corpus; (3) prototype development with corrective agentic logic and hybrid search; and (4) evaluation of technology acceptance and usability based on the Technology Acceptance
Model (TAM). The results evidenced the mitigation of the information gap, ensuring answers anchored in official sources. The investigation further revealed critical normative gaps in per diem and travel processes, demonstrating that automation effectiveness is limited by governance bottlenecks that transcend the scope of artificial intelligence. In usability tests, the prototype proved to be useful and capable of reducing information fragmentation, acting as a support layer for the dissemination of administrative knowledge, with a significant future intention of use among staff members.