Audit Trail
CanonicalConfidence
Cognitive Load
Low
Evidence
production validated
Impact
feature
Live Preview
Agent auto-filled Section 4.2
10:42 AMUser approved Section 4.2
10:43 AMAgent flagged risk in Section 7.1
10:45 AMUser escalated Section 7.1 for review
10:47 AMEthical Guardrail
Every agent action must be logged and reversible for a minimum of 30 days. Audit trail cannot be disabled for safety-critical workflows.
Design Intent
Every action that matters leaves a trace. An Audit Trail gives users complete visibility into what the agent did, when, why, and how -- creating psychological safety and accountability at the same time.
Psychology Principle
Psychological safety and accountability both require traceability and reversibility of AI actions.
Description
A living history log that makes every agent suggestion, decision, and execution fully traceable and reversible.
When to use
Every agent-driven workflow. Non-optional for compliance-sensitive or high-stakes features.
Example
Sidebar Timeline: Chronological entries with plain language, data sources, and one-click Revert this step for every agent action.
Autonomy Compatibility
Behavioral Objective
Users review the audit trail when needed and confidently revert any action.
- Higher willingness to grant autonomy
- Faster debugging and learning from agent behavior
- Full regulatory and psychological safety
Target Actor
role
Everyday user
environment
Compliance-sensitive or high-stakes workflows
emotional baseline
Needs accountability and reversibility
ai familiarity
medium
risk tolerance
low
Execution Model
automatic_logging
Capture every action in real time.
Action occurs without visible log entry.
contextual_access
Make the trail instantly available where the action happened.
User has to hunt for the history.
one_click_revert
Every entry has an immediate undo.
User must manually restore previous state.
Failure Modes
Trail becomes too long and noisy
Group related actions and allow filtering
Entries are too technical
Use plain language and user-friendly summaries
Revert is not truly reversible
Ensure full state restoration
Trail is hidden or hard to access
Keep it persistently visible in context
Privacy concerns with detailed logs
Allow user-controlled retention and anonymization
Agent Decision Protocol
Triggers
- Any action is logged
- User reviews history or questions a change
- High-stakes or autonomous action occurs
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 auto-fills a section -> immediate log entry: 10:42am -- Auto-filled Section X using project history Y. Revert?
Behavioral KPIs
Primary
- Audit trail access rate
- Revert usage rate
- User confidence in agent actions
Risk
- Undetected or un-reversible changes
- Compliance or audit failures
Trust
- User-reported I can always see what the agent did
- Autonomy Dial usage when full trail is present
Decay Monitoring
Revalidate when
- New action types or data sources are added
- Regulatory requirements change
- Trail volume grows significantly
Decay signals
- Users unable to find or understand trail entries
- Rising revert failures
- Feedback that I can't see what the agent changed
Pattern Relationships
Amplifies
Conflicts with