Complete one monitoring plan for each deployed AI system. The plan must be reviewed and signed before the system goes live. Update the plan whenever the system's use case, data inputs, or risk classification changes materially.
1. System Details
System name, version, and risk classification
[System name] · [Version] · Risk: [ Unacceptable / High / Medium / Low ]
2. Monitoring Specifications
| What is Monitored | Monitored By (Role) | Monitoring Method / Tool | Frequency | Alert Threshold | Escalation Path if Breached |
|---|
| Accuracy against benchmark | [e.g. IT Manager] | [e.g. Automated dashboard — [tool name]] | Weekly | [e.g. <92%] | Notify Risk & Compliance → Executive AI Officer |
| Human override rate | [e.g. Business Unit Manager] | [e.g. Monthly report from system logs] | Monthly | [e.g. >20%] | Trigger full system review |
| Data quality of inputs | [e.g. IT Manager] | [e.g. Automated validation rules] | Continuous | [e.g. >5% validation failures] | Suspend AI processing; notify data owner |
| | | | | | |
3. Periodic Review Schedule
| Review Type | Frequency | Responsible | Output |
|---|
| Performance dashboard review | Monthly | [Role] | Updated dashboard; escalation if thresholds breached |
| Full system audit | Annual | [Role] | Audit report; recommendation to continue / modify / decommission |
| Bias re-assessment | Annual | [Role] | Updated bias assessment report |
Plan approved by
[Name] · [Date]
Executive Accountable Official
[Name] · [Date]