Design Principle
AI generation needs to show what the system is doing, what users can still do, and how to recover after failure.
Principle AI generation needs to show what the system is doing, what users can still do, and how to recover after failure.
Design action Generation areas show stage, queue/time, cancel/retry, partial results, and error recovery; avoid a spinner-only state.
Examples Positive example: Text, UI, images, reports, code, and workflows are generated asynchronously by AI. Counterexample: A long task only shows Loading.
Apply when Text, UI, images, reports, code, and workflows are generated asynchronously by AI. Users need to judge state, scope, risk, or next action quickly.
Source notes Source note: Synthesized from GOV.UK, Carbon, Polaris, Atlassian, Design.Systems, Design Systems Repo, and product-pattern libraries.
Agent Directive
Generation areas show stage, queue/time, cancel/retry, partial results, and error recovery; avoid a spinner-only state.