Design and Development of an AI-Powered Intelligent Chatbot for Student Information Services in Higher Educational Institutions Using Large Language Models

Abstract

Higher educational institutions increasingly face pressure to respond quickly and consistently to routine student enquiries on admissions, registration, examinations, course requirements, and institutional policies. This study designed, developed, and evaluated an AI-powered chatbot for student information services using a Retrieval-Augmented Generation (RAG) architecture. Institutional documents, including the student handbook, admission guidelines, frequently asked questions, and course catalogue, were pre-processed into retrievable knowledge chunks and connected to a Large Language Model through a Flask-based backend and browser-based chat interface. A formative pilot evaluation was conducted with 30 volunteer participants, comprising 24 undergraduate students and 6 administrative or academic-affairs staff. Across 120 test queries, 94 responses were rated fully correct, 19 partially correct, and 7 incorrect or irrelevant, giving a combined fully/substantially correct response rate of 94.2%. The chatbot returned responses in an average of 3.8 seconds under normal network conditions, while 82.0% of satisfaction ratings were 4 or 5 on a five-point Likert scale. Only 9 of the 120 queries, representing 7.5%, required fallback or escalation to human staff because they were ambiguous, sensitive, or outside the available knowledge base. The findings indicate that combining Large Language Models with RAG-grounded institutional content can improve the availability, responsiveness, and consistency of routine student information services in Nigerian polytechnic and university contexts, provided the knowledge base is regularly updated and human escalation remains available for complex cases.