PartnerAlly Docs
AI Governance

AI Audit Trail

Track and review all AI decisions and actions in the platform.

AI Audit Trail

The AI Audit Trail provides complete visibility into AI operations within PartnerAlly. It logs what AI did, when, and why—essential for governance, compliance, and debugging.

What the Audit Trail Captures

Every AI action is logged:

Action TypeWhat's Logged
Document AnalysisDocument analyzed, gaps found, confidence scores
Gap DetectionGap created, severity assigned, reasoning
Workflow GenerationWorkflow created, tasks generated, source
Risk PrioritizationPriority calculated, factors used, result
Chat ResponsesQuestion asked, response given, sources used

Accessing the Audit Trail

Go to AI Governance

Navigate to the AI Governance section.

Click "Audit Trail"

Opens the audit log viewer.

Find specific entries.

Audit Log Fields

Each log entry contains:

FieldDescription
TimestampWhen the action occurred
Action TypeWhat type of AI operation
InputWhat was provided to AI
OutputWhat AI produced
ModelWhich AI model was used
ConfidenceAI's confidence level
UserWho triggered the action
ContextRelated items (document, gap, etc.)

Filtering the Audit Trail

By Action Type

Filter to specific actions:

  • Document Analysis
  • Gap Detection
  • Workflow Generation
  • Risk Scoring
  • Chat Assistant

By Date Range

Specify time period:

  • Today
  • This week
  • This month
  • Custom range

By User

See actions triggered by:

  • Specific user
  • System (automated)
  • API

By Confidence Level

Filter by AI confidence:

  • High (80% and above)
  • Medium (50-79%)
  • Low (below 50%)

Filtering by low confidence can help identify AI outputs that may need human review.

Reading Audit Entries

Entry Details

Click any entry to see:

Input Section:

  • What was provided to AI
  • Context available
  • Configuration at the time

Processing Section:

  • Model used
  • Processing time
  • Resources consumed

Output Section:

  • What AI returned
  • Confidence score
  • Reasoning (if available)

Outcome Section:

  • What happened next
  • Human actions taken
  • Whether output was modified

Understanding AI Reasoning

Explainability

For key decisions, AI provides reasoning:

  • Why a gap was identified
  • How severity was determined
  • What factors drove prioritization
  • Why certain tasks were generated

Confidence Breakdown

Confidence scores may include:

  • Contributing factors
  • Uncertainty sources
  • Supporting evidence

Tracking Human Overrides

When Humans Disagree

The audit trail captures:

  • Original AI output
  • Human modification
  • Reason for override (if provided)
  • Final outcome

Override Analysis

Review overrides to:

  • Identify systematic AI issues
  • Improve future recommendations
  • Document human judgment
  • Support audit discussions

Exporting Audit Data

Export Options

FormatUse Case
CSVData analysis
JSONTechnical integration
PDFAudit documentation

Exporting

Apply Filters

Set the scope of data to export.

Click "Export"

Opens export dialog.

Choose Format

Select CSV, JSON, or PDF.

Download

File is generated and downloaded.

Audit Trail for Compliance

Demonstrating AI Oversight

The audit trail helps demonstrate:

  • All AI actions are logged
  • Human review is maintained
  • Overrides are documented
  • Decisions are traceable

Supporting Audits

Provide auditors with:

  • Complete AI activity logs
  • Decision rationale
  • Override history
  • Confidence distributions

Regulatory Requirements

May help satisfy:

  • EU AI Act transparency requirements
  • SOC 2 logging requirements
  • Industry-specific AI governance

Retention and Storage

How Long Data Is Kept

Audit trail retention:

  • Default: 3 years
  • Configurable by organization
  • Follows data retention policies

Data Security

Audit data is:

  • Encrypted at rest
  • Access controlled
  • Immutable (cannot be modified)
  • Backed up regularly

Audit trail entries cannot be deleted or modified. This ensures integrity for compliance purposes.

Monitoring and Alerts

Setting Up Alerts

Create alerts for:

  • Low confidence decisions
  • Unusual AI activity
  • High volume of overrides
  • Error conditions

Monitoring Dashboard

Track AI health via:

  • Activity trends
  • Error rates
  • Confidence distributions
  • Override patterns

Best Practices

Regular Review

  • Review audit trail weekly
  • Look for patterns
  • Identify improvement areas
  • Document findings

Override Documentation

When overriding AI:

  • Provide reason when prompted
  • Be specific about disagreement
  • Document your reasoning
  • Enable continuous improvement

Audit Preparation

Before audits:

  • Familiarize with audit trail
  • Generate relevant reports
  • Prepare explanations
  • Identify human oversight examples

Common Questions

How far back does the trail go?

Depends on retention settings:

  • Minimum 1 year required
  • Default 3 years
  • Maximum based on storage

Can I delete audit entries?

No. Audit entries are immutable to ensure compliance integrity.

Who can access the audit trail?

Access is role-based:

  • Admins: Full access
  • Compliance: View access
  • Users: Own actions only

Is the audit trail included in data exports?

Yes, for organization data exports. User personal data exports include their own actions.

Next Steps

On this page