AI Oversight Settings
Configure human oversight and control settings for AI features.
AI Oversight Settings
Oversight settings control how much human review is required for AI outputs. Configure these to match your organization's risk tolerance and governance requirements.
Why Configure Oversight?
Different organizations need different levels of AI oversight:
| Consideration | Lower Oversight | Higher Oversight |
|---|---|---|
| Speed | Faster AI-driven processes | More review time |
| Control | More AI autonomy | More human control |
| Risk Tolerance | Higher AI trust | Lower AI trust |
| Compliance Needs | Less regulated | Highly regulated |
There's no one-size-fits-all setting. Choose based on your organization's needs and regulatory requirements.
Accessing Oversight Settings
Go to Settings
Navigate to organization settings.
Click "AI Settings"
Opens AI configuration.
Find Oversight Section
Configure oversight options.
Oversight Options
Gap Detection Oversight
Control review requirements for AI-identified gaps:
| Setting | Description |
|---|---|
| Auto-Accept | Gaps created automatically, no review required |
| Require Review | Gaps created as "pending review" |
| Critical Only | Only critical/high gaps require review |
| All Gaps | All AI gaps require human confirmation |
Workflow Generation Oversight
Control workflow creation process:
| Setting | Description |
|---|---|
| Auto-Create Draft | Workflows created as drafts |
| Require Review | Generated workflows need approval before saving |
| Human Edit Required | Workflows cannot be saved without modification |
Severity Assignment Oversight
Control gap severity determination:
| Setting | Description |
|---|---|
| AI Assigns | AI sets severity, user can change |
| AI Suggests | AI suggests, user must confirm |
| User Assigns | AI provides context, user sets severity |
Confidence Thresholds
Set minimum confidence for automatic actions:
| Threshold | Effect |
|---|---|
| High (80%+) | Only high-confidence results proceed automatically |
| Medium (60%+) | Moderate confidence accepted |
| Low (40%+) | Most results accepted |
| None | All results accepted regardless of confidence |
Configuring Settings
Changing a Setting
Select the Setting
Find the setting you want to change.
Choose New Value
Select from available options.
Save Changes
Click save to apply.
Confirm Impact
Acknowledge any warnings.
Recommended Configurations
Conservative (High Oversight)
For highly regulated organizations:
- Gap Detection: All Gaps Require Review
- Workflow Generation: Require Review
- Severity: AI Suggests
- Confidence Threshold: High
Balanced (Default)
For most organizations:
- Gap Detection: Critical Only
- Workflow Generation: Auto-Create Draft
- Severity: AI Assigns
- Confidence Threshold: Medium
Efficient (Lower Oversight)
For faster processing:
- Gap Detection: Auto-Accept
- Workflow Generation: Auto-Create Draft
- Severity: AI Assigns
- Confidence Threshold: Low
Lower oversight settings mean more AI autonomy. Ensure this aligns with your governance requirements.
Review Queue
When Review Is Required
Items requiring review appear in the review queue:
- Pending gaps (if review required)
- Pending workflows (if approval needed)
- Low-confidence items (if threshold set)
Processing the Queue
Open Review Queue
Access from AI Governance or dashboard.
Review Item
See AI output and reasoning.
Take Action
Approve, modify, or reject.
Continue
Process remaining items.
Override Documentation
When You Override AI
Override settings can require:
- Reason selection from dropdown
- Free-text explanation
- Manager approval
- Timestamp and audit log
Documentation Settings
| Setting | Options |
|---|---|
| Override Reason | Required / Optional / None |
| Reason Type | Dropdown / Free Text / Both |
| Approval | Self / Manager / None |
Notification Settings
AI-Related Notifications
Configure alerts for:
- Items requiring review
- Low-confidence outputs
- Override requests
- Queue aging (old items)
Review Reminders
| Setting | Description |
|---|---|
| Daily Digest | Summary of pending reviews |
| Immediate | Alert on each new item |
| None | No notifications |
Role-Based Settings
Admin Settings
Admins can configure:
- Organization-wide defaults
- Role-specific overrides
- Required settings (cannot be changed by users)
User Settings
Users may configure (if permitted):
- Personal review preferences
- Notification preferences
- Display options
Audit and Compliance
Oversight Audit Trail
All settings changes are logged:
- Who changed what
- When changes occurred
- Previous and new values
Compliance Documentation
For audits:
- Current settings report
- Settings history
- Override documentation
- Review queue processing
Best Practices
Finding the Right Balance
- Start with balanced settings
- Monitor for issues
- Adjust based on experience
- Document your rationale
Regular Review
- Review settings quarterly
- After regulatory changes
- When onboarding new frameworks
- Following audits
Team Training
Ensure team understands:
- Current settings and why
- How to process review queue
- When to override AI
- Documentation requirements
Common Questions
Can different users have different settings?
Some settings can be role-based:
- Admins: Lower oversight (trusted)
- Regular users: Standard oversight
- New users: Higher oversight
What happens to pending items if I change settings?
Setting changes affect new items only. Pending items continue under original rules.
Can I require oversight for specific frameworks?
Yes, framework-specific settings may be available. Contact support for configuration.
How do I track oversight effectiveness?
Review:
- Override rates
- Review queue metrics
- Audit trail patterns
- User feedback
Next Steps
- Audit Trail - View AI activity logs
- Model Registry - Understand AI models
- Settings - Other platform settings