Behavioural Engagement Patterns in Mangosuthu University of Technology's AI Student-Support Chatbot

South African universities of technology often handle high volumes of routine admissions, registration and IT-support queries with limited advising staff, and Mangosuthu University of Technology (MUT) in Umlazi, Durban deployed a hybrid AI-assisted chatbot intended to triage these queries automatically and escalate harder cases to human staff.

Researchers from MUT's Learning and Teaching Development Centre and Quality Management Directorate, with a colleague from the Durban University of Technology, set out to analyse chatbot interaction records to understand resolution rates, response times and student engagement patterns at query, user and institutional-team level.

The study analysed 1,495 chatbot interaction records from July to November 2025, with admissions (111 interactions), registration (63) and IT support (31) as the dominant query types, and described three engagement patterns among students: exploratory, transactional and sustained.

The study found a 96.83% resolution rate at the level of individual queries and 89.26% at the level of individual users, but the average first response time was 23 hours and the average overall response time 7.5 hours, and usage was highly concentrated, with one user alone accounting for 360 of 484 user-level interactions.

The authors frame the system as a hybrid human-plus-AI model rather than a fully autonomous one; the long average first-response time is a notable gap between the promise of instant AI support and this deployment's practice, and the concentration of use in a small number of students limits how far the findings generalise across MUT's wider student body.