FREE HUMAN–AI UX TOOL

How much autonomy should this AI action have?

Use consequence and reversibility first. Add model confidence and user expectation second. The output is a product-design starting point, not a safety or legal assessment.

Consequence if wrongReversibilitySystem certaintyDoes the user expect the system to act without asking?

Level 1 — Act automaticallyKeep the action visible and reversible. Avoid unnecessary approval friction.