How Zalando built its own AI prototyping environment on top of its design system
Into Design Systems · 9 November 2026
Read the originalZalando's design system team built its own AI prototyping tool, the ZDS Capsule, after Figma Make stayed off brand, then learned its prototypes could slow implementation.
Dana Baddad and Gonzalo Vásquez Correa of the Zalando Design System team presented the ZDS Capsule at an Into Design Systems meetup in Berlin on 22 October. The tool, built with Claude Code, turns a Figma design or a prompt into a working prototype with real product data. The newsletter Into Design Systems published notes from their talk.
- Before building the tool, the team moved design tokens into their own repository, connected Figma to code components and rewrote visual guidelines as many small design.md files with an index.
- Skills gave different results from run to run, so the team made them run scripts at critical steps. One script returns each icon's real name instead of a screenshot.
- More than 20 designers have contributed to the Capsule and the team has merged 98 of their changes, after small-group support sessions got people past access and setup problems.
- For a kids' wishlist feature, AI-written motion specs confused an engineer, and the team judged implementation could take two or three times as long. It has no answer for handoff yet.
A design system team building its own AI prototyping tool could ask engineers what they need from a prototype before generating specs for them, as the notes suggest. The account is a newsletter's notes from the team's talk.
Split design guidelines into many small markdown files with an index, so an AI agent reads only the file it needs.
When an AI agent's written instructions for a task give different results from run to run, have the agent run a fixed script for the critical step, such as extracting icon names from a Figma file.
Derived by Working Surface from the article; more in the Playbook. Source line: AI makes exploration easy but it does not automatically make collaboration easy.
13 November 2026: AI agents earn more freedom when a team limits what they can reach, from data to design components, and adds a check that does not depend on a glance.