Chris Pollara in Agents
The Executable Design System
A design system used to tell people how things should look. Now software and AI have to build from it.
Before agentic coding, a design system told a designer how something should look. Then it became a shared language between design and engineering. Now, with AI, it has to do even more.
- An agent can query your components.
- It can understand your tokens.
- It can assemble a page.
- It can check whether what it built follows the rules.
We're already seeing this happen. Figma is exposing design systems to agents through MCP. Storybook is doing the same with components, documentation, and testing. The tools are starting to move from describing a system to enabling software to operate on it. That's a pretty fundamental shift.
In our instance at Handoff, an executable design system isn't just documentation for people to read anymore. It has to be something people, software, and AI can build from. And that means the rules can't just exist in a Figma file, a wiki, a code repository, or someone's head.
They need to be structured. Connected. Machine-readable.
Because if AI is going to help build your digital experiences, it needs more than instructions. It needs to understand the system.