Generative AI is rapidly entering university classrooms, yet many educators remain uncertain about how to move beyond using AI as a general-purpose chatbot or productivity aid. When AI is introduced without pedagogical design, it can unintentionally reduce student thinking, over-scaffold learning, or undermine intended outcomes. This hands-on workshop addresses this challenge by guiding educators through a pedagogy-first process for both designing and building AI learning agents.
The workshop begins by reframing AI agents as pedagogical artefacts, not tools. Participants will learn how to identify authentic learning problems, specify the target thinking they want students to develop, and select instructional strategies that legitimately benefit from AI support. Using a structured Faculty Design Input Template, participants will articulate the agent’s pedagogical role, boundaries, guardrails, conversation control logic, and fading mechanisms to ensure students ultimately perform independently.
Crucially, the workshop goes beyond design. Participants will translate their completed templates into actual system prompts and use them to build and test a working AI agent during the session. Through guided testing, participants will observe how different design choices affect agent behaviour, including when the agent should probe, redirect, summarise, or end an interaction. Common pitfalls—such as over-helping, answer dumping and dependency—are surfaced and corrected through live iteration.
By the end of the session, participants will leave not only with a pedagogically sound design blueprint, but also with a functional AI learning agent that is aligned to their teaching context and ready for classroom use or further refinement.
Key takeaways include:
Join our EDUtech community and access our exciting library of free-to-watch edtech videos.