Error rate describes how often an AI system produces incorrect outputs. Communicating it concretely helps users stay alert without treating the system as either infallible or unusable. section 12.3.2
Demand forecasting under changing conditions
A forecasting platform may exceed 95% accuracy for stable demand patterns, while weather shocks or geopolitical events slightly increase its error rate. Planners should compare predictions with local knowledge, adjust mismatches manually, and flag problems for review. section 12.3.2
Make error rates useful by pairing them with controls:
- Explain when performance changes and identify low-confidence or anomalous outputs. section 12.3.2
- Provide manual overrides and simple reporting so user feedback can improve later updates. section 12.3.2
