A tall wooden card catalogue with many small drawers, where a mechanical arm reaching into one empty drawer stops and raises a small flag instead of placing a card, while a figure seen from behind walks over to look, the whole scene drawn with wide empty margins on every side.

Design Systems Collective

Your Design System Isn’t AI-Ready Yet

Shadi Abd · 24 October 2026

How-to · design systems · agents · Agent context files

Takeaway

Shadi Abd argues that an AI agent working in an enterprise design system guesses wherever rules are unwritten, so teams should record the rules and verify them with tests.

Summary

Shadi Abd, writing in Design Systems Collective, a publication about design systems, describes a large company whose design system serves several products built in different code frameworks. When an AI agent builds screens there, it brings back retired patterns and rebuilds components that already exist. Abd traces the cause to knowledge that designers and developers hold in their heads and never wrote down.

Key points

Abd proposes four layers: define each component and token, map each design part to its code, record which rules apply to which product, and verify the output with tests.
Instructions for AI agents should run from company rules to design-system rules to product rules to the task, with detail retrieved only when it applies.
Where no approved pattern or documented exception exists, the agent should not create a new pattern but should report the gap for review.
Linting, accessibility tests and other automated checks enforce the rules; written instructions only explain them.

Implication

A design-system team can start with an inventory of its most used components and a written rule for what an AI agent does when nothing fits. The piece is one practitioner's proposal, not a report of results.

Plays

Design

Write beside each design-system component's properties when to use it and when not to, so an AI agent can read the reasoning as well as the options.

Design

Instruct AI agents that when no approved component, pattern or documented exception fits, they report the gap for review instead of creating a new pattern.

Derived by Working Surface from the article; more in the Playbook. Source line: Where the system is silent, the agent will infer.

Source issue

25 October 2026: AI agents and people outside engineering can change real products safely once a team has built shared templates, written rules and automated checks.