Agentic Design System

Part 5 and the capstone of the Agentic Design Systems series. Tools and loops are not an operating model. This is how the people, the process, and the governance fit together so a team of humans and a fleet of agents can run a design system sustainably, without the whole thing depending on the one person who understands the agents.

Part 4 of the Agentic Design Systems series. When an agent has earned the right to fix things, the detect-propose-verify-merge-learn loop lets your design system repair its own drift while you sleep. Here is the loop, the four guardrails that keep it safe, and the moment you should switch it off.

Part 3 of the Agentic Design Systems series. Observer is the longest, cheapest, most-skipped trust level. Run an agent read-only against your real system for weeks before you give it any authority, and let the audit decide whether it should ever be promoted at all.

Part 2 of the Agentic Design Systems series. The five trust levels (Observer, Advisor, Junior, Senior, Autonomous), what each one is allowed to do, and the exact criteria to promote an agent from one level to the next.

Part 1 of the Agentic Design Systems series, based on Romina's IDS 2026 talk. A design system built for AI agents, not just humans. Stops hallucination before it ships.