Design Systems Are Becoming the Control Layer for Agents
Why this matters to Wen: once AI starts helping with implementation, sloppy systems stop being an internal annoyance and start becoming a quality liability.
Evidence: Decagon says its Figma library now spans hundreds of components and uses Figma MCP plus Figma Make to keep design, code, and agents inside the same loop.
The Story
Figma’s Decagon case study is less about one company’s tooling stack and more about what design systems are turning into. They are no longer just tidy repositories for consistency-minded teams. They are becoming the control layer that tells coding agents what “correct” actually looks like.
That shift matters because agents do not improvise like a strong product designer or engineer can. They work from whatever structure, naming, states, and component logic you give them. Decagon’s story is compelling precisely because it treats the design system as operational infrastructure. The team built parity between Figma and Storybook, created agent skills to pull exact components, and reduced the amount of interpretive drift between mockup and shipped work.
For Studio Wensday, the lesson is bigger than software teams. Any repeatable brand system, content format, or page pattern that you want AI to help with needs stronger bones first. Good taste still matters, but it starts becoming more powerful when it is captured in reusable decisions instead of left floating in your head.
Studio Wensday Angle
Pick one recurring asset type, like a carousel cover, newsletter section, or landing-page CTA block. Document the exact states, spacing rules, and tone rules you actually want, then test whether another person or a tool can reproduce it cleanly from that system.