Reversible Actions
CanonicalConfidence
Cognitive Load
Low
Evidence
production validated
Impact
feature
Ethical Guardrail
Never perform irreversible actions without explicit user confirmation. Never hide the undo mechanism. Always default to draft/suggest mode for high-stakes autonomous actions.
Design Intent
Every irreversible action creates anxiety. The brain treats potential mistakes as threats, so users hesitate or avoid actions altogether. Reversible Actions makes every meaningful change undoable with one click.
Psychology Principle
Every irreversible action creates anxiety. The brain treats potential mistakes as threats.
Description
Make every meaningful change undoable with one click, giving users psychological safety to experiment and trust the system.
When to use
Every action that could alter data, state, or outcomes -- especially in AI-driven workflows where the agent acts on behalf of the user.
Example
Figma Unlimited Undo: Every change is instantly reversible with Cmd+Z -- users freely experiment knowing nothing is lost.
Autonomy Compatibility
Behavioral Objective
Users confidently take actions knowing they can instantly undo them if needed.
- Higher experimentation and feature adoption
- Reduced anxiety in decision-making flows
- Increased trust in AI-driven automation
Target Actor
role
Everyday user
environment
High-stakes or unfamiliar workflows
emotional baseline
Fear of permanent mistakes
ai familiarity
medium
risk tolerance
low
Execution Model
pre_action_signaling
Clearly communicate that the action is reversible before it happens.
User hesitates or asks Is this permanent?
instant_undo
Provide one-click reversal immediately after the action.
User must navigate elsewhere or perform multiple steps to undo.
history_visibility
Show what was changed and allow selective restoration.
User doesn't know what changed or how to get back.
Failure Modes
Undo window is too short
Default to at least 30 seconds or 30 days for critical data
Undo feels hidden or complex
Surface it prominently in context
Over-reliance on undo encourages careless actions
Pair with gentle confirmation for very high-stakes changes
Technical limitations make true reversibility hard
Fall back to draft mode and clear warnings
User forgets the action was reversible
Remind them once via agent or notification
Agent Decision Protocol
Triggers
- User hesitates before a potentially destructive action
- Feature involves data changes or AI automation
- User asks What if I change my mind?
Escalation Strategy
L1: Diagnose the failing element via behavioral_signals
L2: Nudge -- adjust copy, timing, or visual salience
L3: Restructure -- simplify flow, add progressive disclosure, restructure form
L4: Constrain -- lock Autonomy Dial to confirm_execution, add Strategic Friction
L5: Yield -- flag for human designer or domain expert review
Example
Agent suggests deleting an old file -> shows This can be restored from trash anytime + one-tap undo after deletion.
Behavioral KPIs
Primary
- Action completion rate when reversibility is signaled
- Undo usage rate
- User confidence score before/after action
Risk
- Abandonment due to fear of permanence
- Permanent-error incidents
Trust
- User-reported I feel safe trying things
- Autonomy Dial usage when agent performs reversible actions
Decay Monitoring
Revalidate when
- New destructive actions or AI capabilities are added
- Data retention policies change
- User expectations for undo windows evolve
Decay signals
- Rising pre-action hesitation
- Increased support tickets about I can't undo this
- Drop in experimentation rates
Pattern Relationships
Supports
Amplifies
Conflicts with