University websites are full of valuable content, yet serving the needs of diverse audiences - from school-leavers to seasoned academics - remains a major challenge. How do we present the right information, in the right way, for everyone? And how can hundreds of content editors maintain consistency across such a digital ecosystem? 

At the University of Edinburgh, we teamed up with open-source Drupal AI experts to explore how AI-powered tools, particularly LLMs, could support smarter, more sustainable content management. Using a UX-led research approach, we ran a series of controlled experiments to test what AI could do, and what it's limitations could be. 

What we found was both surprising and exciting. AI agents that teach content editors how to use publishing tools in real-time? Prompt engineering to embed style guide rules into content workflows? AI-generated embeddings to enable more intuitive search results? All of these ideas are showing real promise. 

In this session, I'll share our iterative experimental framework, demonstrate our approaches and offer tips on designing your own AI experiments to investigate potential for your websites. Expect insights from conversational and learning design, interaction and service design, and a healthy dose of AI curiosity. 

Audience: All Track(s): AI, Content Strategy, UX / Design