AWS Launches New AI Agents

Image on Unsplash

Amazon Web Services announced at its re:Invent conference on Tuesday that it had developed a new class of AI agents called Frontier Agents. They are capable of working autonomously and at scale for extended periods, meaning they can operate for hours or days without any human intervention whatsoever.

The company unveiled three specific frontier agents that represent different use cases and capabilities. The first is an agent called Kiro that functions as a virtual developer and can write and test and deploy code. The second is an AWS Security Agent that acts as a security consultant by identifying vulnerabilities and recommending mitigations. The third is an AWS DevOps Agent that functions like an on-call operational team and can diagnose and resolve infrastructure issues.

What separates these agents from previous versions of AI automation systems is that they can reason through complex problems and make decisions autonomously without requiring step-by-step human guidance. Organizations can deploy them for tasks that previously required constant monitoring.

AWS also introduced Amazon Bedrock AgentCore, which it described as the most advanced platform for securely building and deploying agents at scale.  The company emphasized that while AI agents can reason and act autonomously, there is a need for robust controls to prevent unauthorized actions.

AWS also announced new computing infrastructure to support these agents, including EC2 Trn3 UltraServers powered by the advanced Trainium3 chip. The company said they offer 4.4 times greater compute performance and 4 times greater energy efficiency compared to their predecessors.

What these announcements suggest is that the race to deploy autonomous agentic systems at enterprise scale is accelerating. Companies that can build reliable infrastructure and governance systems around these agents will likely capture significant market share.