AI in Higher Education: From Literature to a Course-Anchored Chatbot

Artificial intelligence is reshaping higher education by enabling personalized support for students and efficiency gains for faculty. Yet, meaningful adoption requires more than technology: it demands alignment with pedagogy, careful governance of data and sources, and demonstrable educational value. This thesis investigates how AI can support university-level teaching and learning along two axes: student support and faculty support; then translates the resulting design principles into a working, course-specific AI assistant for the Analysis and Management of Production Systems (AMPS) course. A structured research methodology (Chapter 3) was applied to build the theoretical foundation (Chapter 2) and the cross-case synthesis of applications (Chapter 4), drawing primarily on Scopus-indexed literature with explicit inclusion/exclusion criteria and coding. These insights informed the design of an AMPS Chatbot (Ch.5) developed in Microsoft 365 Copilot Studio and grounded in a closed corpus. The assistant is constrained to: (i) answer only from the provided corpus; (ii) preserve course notation and terminology; and (iii) produce standardized slide-file citations (page ranges on demand). A test methodology and KPI dashboard assess accuracy, completeness, citation compliance, pages-on-demand behavior, style adherence, and out-of-scope guardrails. The prototype demonstrates that a closed-corpus, prompt-engineered assistant can deliver pedagogically consistent explanations with robust traceability to official materials, offering just-in-time support that complements instruction. KPI-based evaluation shows strong citation discipline and style adherence, solid correctness/coverage on in-scope questions, and reliable refusal behavior on out-of-scope prompts, while highlighting improvement areas for cross-module synthesis and page-precision in specific cases.