Amazon Bedrock AgentCore has rolled out new features aimed at enhancing the deployment of AI agents in various industries. In just five months since its preview, the AgentCore SDK has seen over 2 million downloads. Companies like PGA TOUR have harnessed this technology to boost content generation efficiency by 1,000% while cutting costs by 95%. Workday has utilized AgentCore for its Planning Agent to streamline financial data analysis, reducing routine planning analysis time by 30%, equating to about 100 hours saved each month. Grupo Elfa leverages AgentCore Observability for real-time metrics and full audit traceability, significantly improving their operational processes. The latest updates focus on maintaining control over agent actions while allowing for their autonomous operation. The new policy feature in AgentCore sets clear boundaries for agent actions, ensuring compliance and safety. Users can create policies through natural language or Cedar, an open-source policy language, making it accessible for teams without coding expertise. This allows organizations to define exactly what tools and data agents can access and under what conditions. The policy engine can be tested in a logging mode before full deployment, ensuring that organizations can validate their rules without risking operational integrity. AgentCore Evaluations introduces a managed service for ongoing performance monitoring of agents. This feature assesses agents based on real-world behavior across various quality dimensions, such as correctness and context relevance. Users can implement built-in evaluators or design custom scoring systems tailored to specific business needs. These tools aim to help organizations confidently deploy AI agents while ensuring they meet operational standards and maintain high-quality interactions.