Design and Evaluation of an AI-Powered Chatbot for ERP-Based Student Services in Higher Education

Higher education institutions are increasingly exploring conversational AI to enhance student services, especially in STEM programs where timely support is crucial. This paper presents the design and evaluation of a practical AI-powered chatbot prototype that assists with essential student service functions by accessing ERP-style data. The proposed chatbot leverages natural language processing (NLP) to handle queries related to admissions, course management, and academic advising, providing instant, and round-the-clock conversational support. A prototype was developed in Python and evaluated on key metrics: query resolution accuracy, system response time, task completion rate, and satisfaction score. To provide interpretable and institution-friendly automation, we adopted a rule-based approach using keyword recognition. This study utilized a stimulated data set of 100 student records, and chatbot was tested across 50 queries. The chatbot achieved an accuracy rate of $100 \%$, a 0.001 second average response time, and an $100 \%$ task completion rate, outperforming traditional support systems ($80 \%$ accuracy, 45-second response time). The results show that stimulated user satisfaction reached 4.5/5, indicating measurable enhancement in the student engagement and satisfaction. These results demonstrate that an ERP-integrated chatbot can significantly benefit STEM education by providing on-demand, tailored assistance, reducing staff workload, and supporting the modernization of academic service delivery