LLMs in government service design
There are incredibly effective principles and patterns for government service and content design. However, complex services like statutory planning have historically resisted these same principles, patterns (and hard work) that have made other digitised services feel simple.
At myLot, we’ve made headway by using large-language models (LLMs) to deal with the immense number of variables that come with a seemingly straightforward question like “can I build a shed in my backyard?”, but it’s brought with it a host of new design challenges and nuanced trade-offs.
What about statutory planning justifies the use of LLMs? How do we make the affordances of LLM interactions clear, while still leveraging government service design patterns that we know to work? How do we design good content when it’s all generated by an LLM? And how do we make sure that content is not only clear, but also accurate?
Harry Court
Harry Court is a UX designer based in Eora (Sydney) with a practice that primarily focuses on content design and information architecture. He currently works at myLot, an Australian start-up that makes software for the urban planning industry. For a lot of his career he’s worked in the digital civic design space, where the axiom of “content is the user experience” rings loudly and has been a strong influence on his practice.