AWS New AI Agent Builder Features: Policy, Evaluations & Memory Revealed at re:Invent (2025)

AWS reveals groundbreaking AI agent advancements, sparking industry excitement and debate.

Amazon Web Services (AWS) is revolutionizing the AI agent game with its latest enhancements to Amazon Bedrock AgentCore. The platform's new capabilities aim to simplify the creation and management of AI agents for businesses, addressing key challenges in the rapidly evolving AI landscape.

At the AWS re:Invent conference, AWS unveiled a trio of powerful AgentCore features. First, the Policy feature empowers developers to define interaction boundaries for AI agents using natural language. This innovation seamlessly integrates with AgentCore Gateway, ensuring each agent's actions adhere to specified rules, automatically halting any violations. For instance, developers can now restrict access to sensitive data or third-party apps like Salesforce or Slack, and even set refund limits for AI agents, requiring human intervention for larger amounts.

But here's where it gets controversial: AWS also introduced AgentCore Evaluations, a collection of 13 evaluation systems that assess AI agents on various metrics, including accuracy, safety, and tool selection. This suite promises to alleviate developer concerns about agent deployment, but will it live up to the hype? It's a much-needed tool, but building it from scratch can be a daunting task.

Additionally, AWS is adding AgentCore Memory, enabling agents to record user information, such as travel preferences, and leverage this data for more informed decisions. This feature could be a game-changer for personalized AI experiences.

"We're enhancing AgentCore's core functionalities to make AI agents more robust and user-friendly," said David Richardson, vice president of AgentCore. "Policy ensures controlled interactions, Memory enhances decision-making, and Evaluations provide a safety net for developers."

As AI agents gain traction, opinions vary on their longevity. Richardson believes AgentCore's tools are future-proof, adapting to shifting market trends. But is this optimism justified? Will AI agents continue to dominate, or will they fade like a passing fad?

Stay tuned for TechCrunch's ongoing coverage of this exciting tech event, and join the conversation! What's your take on the future of AI agents? Are AWS's new features the key to their success, or is there more to the story?

AWS New AI Agent Builder Features: Policy, Evaluations & Memory Revealed at re:Invent (2025)
Top Articles
Latest Posts
Recommended Articles
Article information

Author: Trent Wehner

Last Updated:

Views: 5944

Rating: 4.6 / 5 (76 voted)

Reviews: 83% of readers found this page helpful

Author information

Name: Trent Wehner

Birthday: 1993-03-14

Address: 872 Kevin Squares, New Codyville, AK 01785-0416

Phone: +18698800304764

Job: Senior Farming Developer

Hobby: Paintball, Calligraphy, Hunting, Flying disc, Lapidary, Rafting, Inline skating

Introduction: My name is Trent Wehner, I am a talented, brainy, zealous, light, funny, gleaming, attractive person who loves writing and wants to share my knowledge and understanding with you.