I gave this lightning talk during AI Week, organised by the Center for Innovation in Teaching and Learning (CITL) at the American University of Sharjah. The premise is simple: AI now completes most take-home assignments at a level indistinguishable from student work, and detection tools are unreliable. Rather than fighting this, we can redesign assessments to use AI as a learning tool while still measuring genuine understanding.

I walk through what is actually happening under the hood (next-token prediction, and why that is “no understanding, just sophisticated pattern matching”), the evidence that both students and teachers are already using these tools heavily, and a three-level AI usage policy framework (No AI, Guided AI, and Full AI Integration) that instructors can map to their own course learning outcomes.