This paper examines the use of generative artificial intelligence, specifically ChatGPT, to support the evaluation of student assignments and feedback in a project-based learning (PBL) Business Continuity Management (BCM) module across the 2023 and 2024 academic years. The study compares AI-generated analyses with traditional human assessment to evaluate consistency, depth, and reliability. Multiple evaluation rounds using different prompts and users show that ChatGPT can efficiently identify themes, extract patterns, and apply rubric-based criteria, offering meaningful support for processing large volumes of qualitative work. However, limitations arise around contextual nuance, over-generalisation, and prompt sensitivity, highlighting the continued need for human oversight. The study also demonstrates how AI can contribute to rubric refinement by surfacing overlooked elements and misalignments between learning outcomes and assessment criteria. Practical principles for responsible and transparent integration of AI into assessment are proposed, with implications for teaching practice, feedback quality, and digital responsibility in higher education.