AWS Launches Agent Quality Optimization in AgentCore — Production Traces to A/B Testing Pipeline

AWS Machine Learning Blog·medium signal

AWS announced agent quality optimization in AgentCore (preview), enabling teams to generate recommendations from production traces, validate them with batch evaluation and A/B testing, and ship agent improvements with confidence. This addresses the 'day 2' problem of agent deployment — agents that perform well at launch degrade over time as models and data drift. For builders running production agents, this is the first major cloud-native agent observability-to-improvement pipeline.