Across higher education, institutions are adopting generative AI to accelerate curriculum design, yet many approaches risk over-automation or fragmented use. This paper presents an AI-assisted course development model implemented at UNSW College that embeds human judgement, governance, and pedagogical integrity at every stage. Rather than treating AI as an autonomous content generator, the model positions it as a coordinated set of specialised agents operating under human stewardship.

The AI Course Developer functions as a staged workflow covering course review, curriculum architecture, assessment design, learning content development, and resource production. Each stage is supported by a purpose-built AI agent with focused expertise, such as curriculum alignment, rubric design, or assessment risk analysis. Progression between stages is gated by human review. This human-in-the-loop approach mitigates hallucination risk and ensures alignment with institutional standards, industry relevance and disciplinary judgement.

Scalability is supported through lightweight project management and information architecture. A master control document tracks courses across development stages, while a standardised folder structure and centralised policy knowledge base promote consistency and reduce prompt complexity.

The development outlined a pilot-first implementation in which two courses from different disciplines were run end-to-end through the workflow to identify bottlenecks, refine prompts, and calibrate the human review effort. Early outcomes indicate reduced development time while maintaining constructive alignment, assessment authenticity, and Universal Design for Learning principles. The case study offers practical insights for institutions seeking governed, human-centred AI systems that enhance academic expertise.