The actor triggers something and the system needs to be sure they mean it. Inline confirmation keeps the question where the intent was expressed: the triggering control transforms in place into a confirm/cancel pair, and only a second, deliberate activation completes the act. Because the confirm affordance appears where the actor is already looking and clicking, the friction added is one decision, not a context switch — enough to interrupt autopilot, not enough to interrupt work.
Behaviours
- Timeout reversion. If the actor doesn’t confirm within a few seconds, the control reverts to its resting state — an abandoned near-miss should leave no armed control behind.
- Explicit cancel. Offer a cancel affordance alongside confirm rather than relying on the timeout alone; changing your mind should not require waiting.
- Distinct confirm state. The confirming control must look visibly armed (colour, icon fill) so the actor can tell a loaded control from an idle one.
Choosing against the alternatives
Inline confirmation sits between two cheaper and one heavier option. When the act reverses cleanly, skip confirmation entirely and offer undo with transient feedback — recovery after beats interrogation before. When consequences are severe or affect others, escalate to a dialog that can spell them out. Inline confirmation covers the middle ground.
Related patterns
Alternatives
- Undo — moves the safety net after the act instead of before it; prefer undo when the action reverses cleanly
Related
- Action consequences — the framework that places this move — moderate recovery effort, clear intent, context worth keeping; one rung above undo-only, one below modal
- Deletion — the action it most commonly guards; deletion's decision tree recommends it for items that take minutes, not seconds, to recreate
