AI Change Management
Controlled processes for modifying AI systems while maintaining compliance and performance.
Definition
AI Change Management encompasses processes for safely modifying AI systems in production. AI changes present unique risks because model updates can have unpredictable effects on behavior.
Key AI change management considerations:
- Model retraining: Updates to model weights based on new data
- Feature changes: Adding, removing, or modifying input features
- Threshold adjustments: Changing decision boundaries or confidence thresholds
- Infrastructure changes: Updates to serving infrastructure affecting AI behavior
- Vendor updates: Third-party AI API or model version changes
Best practices include: staged rollouts, A/B testing, rollback capabilities, impact assessment before changes, documentation of all changes, and re-validation after significant updates.
Sources
- •NIST AI RMF
- •ITIL Standards
Related Terms
Risk Management
Systematic process of managing AI-related risks....
AI Monitoring (Production)
Ongoing surveillance of AI systems in production to detect drift, failures, and compliance issues....
AI system operation and management
Deployment, monitoring and maintenance of operational AI models....
Vendor Risk Management (AI)
Framework for assessing and managing risks from third-party AI service providers....
Model Drift
Degradation or change in model performance over time....