Amazon AWS Steps Forward on Agentic AI
By Jim Lundy
Amazon AWS Steps Forward on Agentic AI
Amazon AWS is finding its Generative AI groove. Generative AI continues to shift enterprise application development into a new paradigm. AWS re:Invent 2025 delivered a series of announcements indicating Amazon’s clear focus on moving AI agents from prototypes to production environments. This blog, part I of a three part series on AWS re:Invent, overviews the new AWS AI agents and Bedrock tools and offers our analysis.
Why did AWS announce Frontier Agents and Nova 2?
The new Amazon Nova 2 model family—designed for superior price-performance across reasoning, code generation, and agentic tasks—lays the foundation for more sophisticated autonomous systems. Building on this base, AWS unveiled three new frontier agents: Kiro (virtual developer), AWS DevOps Agent (operational team), and the AWS Security Agent (security consultant).
These agents are designed to execute complex tasks over extended periods, a crucial capability for real-world enterprise automation. Complementing these releases, Bedrock AgentCore added Policy, Evaluation, and Memory capabilities, providing developers with the controls necessary for deploying production-ready agents at scale.
Analysis – Developer focus
It is no surprise that Amazon AWS is focusing on developers – partly because they want enterprises to develop and run their services on AWS. This move to expand in Agentic AI confirms that autonomous AI execution is the next major battleground for hyperscalers. The release of the three domain-specific agents is a solid move, but compared to many other firms, this is late to market. However, the timing of the AWS Security Agent reveals a competitive gap.
AWS: Late on Security Agents
Microsoft has aggressively pushed its Copilot for Security since March 2023, with regular updates. This has allowd Microsoft to embedd its security agent deep within its enterprise ecosystem, creating strong incumbency. While the AWS Security Agent addresses a critical need, Amazon is playing catch-up in what is arguably the most sensitive enterprise use case. This delay means AWS will have to fight harder for mindshare and budget in the high-stakes security operations market against an established competitor.
What should Enterprises do?
Enterprises must recognize that the agent development infrastructure—not just the agents themselves—is now the critical technology stack. The new Bedrock AgentCore features, especially the ability to set Policy and perform continuous Evaluation, transform agent governance from an afterthought into a managed capability. Do not rush to deploy a specific AWS frontier agent; instead, evaluate AgentCore and its controls. Consider how its architecture facilitates integration with non-AWS security frameworks, given Microsoft’s early lead in that specific vertical.
One thing for enterprises to evaluate- Amazon pushing Agents for development and other firms pushing Agents for operational use. Aragon recommends looking for configurable agents – not just ones a provider is offering you to entice your developers to continue using AWS infrastructure.
Bottom Line
AWS has delivered the necessary platform and specialized agents to propel autonomous AI into the enterprise mainstream. The combination of Nova 2 models and AgentCore provides a robust path for development teams. Enterprises should evaluate AgentCore immediately, particularly its Policy and Evaluation features, as they represent the foundation for future governance. While the security agent is a major step, organizations must proceed with careful due diligence to understand the competitive gap against Microsoft and its impact on existing security tooling.
Have a Comment on this?