AWS has opened a public preview of Well-Architected Agent, a generative AI service that analyses AWS environments and recommends ways to improve cost, security, performance and resilience. Its practical pitch is more than a list of findings: AWS says it can suggest fixes tailored to a customer’s goals and environment, including infrastructure-as-code changes.
AWS AI Watch analysis
What happened
Customers create an agent profile to choose which accounts, applications and Regions the service can inspect, which optimisation areas to focus on, and what goals to prioritise. They must also provide customer-managed IAM roles for the agent to read resource configurations, utilisation metrics and application topology. AWS says recommendations should appear within 24 hours of profile creation.
The preview can also review pre-deployment infrastructure projects uploaded as Terraform, CloudFormation or CDK files. Findings range from individual resources to application-wide issues and architectural patterns. Suggested remediation can take the form of console steps, updated infrastructure-as-code templates or AWS CLI commands. Recommendations are available through the console and API.
The service is available in US East (N. Virginia), US East (Ohio) and US West (Oregon), for customers with an AWS Support plan. Workloads can be onboarded from any commercial AWS Region. AWS says generative AI recommendations may contain errors or omissions, and customers remain responsible for evaluating them and applying appropriate oversight. AWS’s preview announcement has the setup details.
Why it matters
Cloud architecture reviews can mean trawling through services, metrics and checklists before getting to a useful decision. AWS is trying to turn that work into prioritised, environment-specific recommendations, with implementation options close at hand. For teams already on AWS, the prospect of getting from a finding to a proposed code change without stitching together several tools is the useful part.
It also puts an AI system in a position to inspect operational infrastructure and suggest changes affecting security, cost and resilience. That makes the permission setup and human review more than box-ticking: a plausible-looking fix still needs to fit the workload it is meant to improve.
Our read
This is a substantial cloud-operations feature, not just another chatbot bolted onto a console. The combination of workload context, pre-deployment reviews and fixes in familiar formats gives it a clear job to do. AWS’s own warning about errors is worth taking seriously; treat the agent as a reviewer that can speed up the investigation, not as an architect whose suggestions ship themselves. For teams eligible for the preview, the sensible test is a scoped profile and a human review of every proposed change.
What to watch
- Whether recommendations prove useful across different workload types, rather than mainly in AWS’s examples.
- How much detail the agent gives about the evidence and trade-offs behind each proposed fix.
- Whether customers can readily assess the quality of its infrastructure-as-code changes before deployment.
Discussion spark: Would you let an AI agent propose changes to live cloud architecture if a person must approve them, or should these tools be limited to pre-deployment reviews?
Sources and evidence
- Announcing AWS Well-Architected Agent, an AI-powered intelligence to optimize your cloud environment (preview) (1 October 2026, 20:04 UTC)
not affiliated with or endorsed by Amazon Web Services (AWS)