AI KRI Wizard
The AI KRI Wizard lets you describe what you want to monitor in natural language. The AI translates your description into a fully configured KRI with query definition, thresholds, and scheduling recommendations.
How It Works
- Navigate to Monitoring → KRIs → + New KRI → Use AI Wizard
- Describe what you want to monitor in plain language
- The AI generates a complete KRI configuration
- Review, adjust, and save
Example Prompts
| Prompt | Generated KRI |
|---|---|
| "Monitor the number of overdue action plans across all engagements" | Query: COUNT(action_plans WHERE status = OVERDUE), Threshold: Warning at 5, Critical at 10, Frequency: Weekly |
| "Track average time to close findings by severity" | Query: AVG(days_to_close) GROUP BY severity, Thresholds per severity level, Frequency: Monthly |
| "Alert when any business unit hasn't been audited in 18 months" | Query: Universe items not audited in 18+ months, Threshold: Any match = Warning, Frequency: Monthly |
| "Monitor the compliance score trend for NCA ECC framework" | Query: Compliance score for NCA ECC over time, Threshold: Warning below 80%, Critical below 60%, Frequency: Quarterly |
What the AI Generates
For each prompt, the wizard produces:
| Component | Description |
|---|---|
| Title | A descriptive KRI name |
| Description | What the KRI monitors |
| Query Definition | SQL or logic to compute the metric |
| Category | OPERATIONAL, FINANCIAL, COMPLIANCE, or STRATEGIC |
| Thresholds | Warning and critical levels with reasoning |
| Frequency | Recommended monitoring frequency |
| Universe Link | Suggested audit universe item connection |
| Risk Link | Suggested risk library item connection |
Review & Customize
After the AI generates the KRI:
- Review the query — ensure it matches your intent
- Adjust thresholds — fine-tune for your risk appetite
- Set the schedule — accept or modify the suggested frequency
- Link entities — connect to relevant universe/risk items
- Save as Draft — the KRI enters DRAFT status for validation
AI Governance
The AI Wizard produces a suggestion. The generated KRI is saved in DRAFT status and must be validated by QA and published by CAE before it becomes active. All AI interactions are logged in the audit trail.
Wizard vs. Playground
| Feature | AI Wizard | KRI Playground |
|---|---|---|
| Natural language input | ✅ | ❌ |
| Visual pipeline builder | ❌ | ✅ |
| AI-generated query | ✅ | ❌ |
| Manual SQL writing | ❌ | ✅ |
| Drag-and-drop | ❌ | ✅ |
| Best for | Quick creation | Precise control |
Both methods produce the same KRI object — they're complementary approaches.
Permissions
The AI KRI Wizard requires:
kri:createpermission (QA, Manager, CAE)ai:suggestpermission (for AI interaction)- An active AI model configured in Administration → AI Models