Architecture documentation is often either too vague to be helpful or too detailed to stay relevant. And let’s be honest, most of us avoid writing it until we have to. But with the right AI tools, we can integrate documentation and design decisions into our workflow without slowing down.

In this live coding session, I’ll demonstrate how GitHub Copilot’s Agent Mode (and LLMs in general) can actively support architectural work, from capturing business drivers and quality attributes to comparing design alternatives and documenting decisions in ADRs.

- We’ll start with an unstructured requirements document, extract the key points, and build a utility tree to guide our priorities.
- From there, we’ll explore architectural options using AI suggestions, make trade-offs visible in comparison tables, and validate our choices against the original goals.
- To wrap it up, we’ll scaffold a cloud-native application based on the selected architecture.

This session is for tech leads and architects who want to bring architecture closer to code and make it more practical using today’s AI tooling.