Every AI design discussion meant re-explaining the product
Using AI to assist with design meant starting from zero every time. It didn't know the product logic, the component rules, or the UI patterns the team already relied on, and the proposals it gave often drifted from how the product actually worked.
Instead of re-explaining the same context over and over, I started documenting that knowledge in a structured format AI could read on its own before joining the discussion.
Documenting the Knowledge, Step by Step
The knowledge base has three parts:
wiki
Business logic, process rules, and other product knowledge
design-system
Visual tokens, component specs, and design patterns
design
New feature ideas and past design decisions
Building it also meant building a few matching Skills: interviewing to capture knowledge, navigating the knowledge base, turning component specs into standardized docs, and guiding design exploration.
The product knowledge layer is fully built. I'm now working through the design system's component specs and pattern inventory, and the next step is experimenting with having AI take a more active role in producing design output 🧪
This page is a living document and will keep updating as the project progresses.