AWS opened a preview of the Well-Architected Agent, an AI service that reviews a cloud environment against Well-Architected best practices and returns prioritised recommendations across cost, security, resilience and performance.
The agent correlates resource configurations, utilisation metrics and application topology, then analyses them against guidance spanning more than 65 AWS services. Every finding arrives with an implementation package, and the agent surfaces trade-offs across pillars rather than optimising one in isolation.
Setup assumes goals were written down
Users start by defining an agent profile with accounts, regions, optimisation pillars and business goals. The agent scans those accounts and ranks recommendations by impact and effort at resource, application and architecture level, including cost impact and infrastructure-as-code changes.
API access lets teams push findings into existing development and operations workflows, and the AWS MCP Server connects the agent to AI coding tools.
Practitioners raised the obvious objection. Nicolas Payette, chief executive of Specira AI, asked where the agent reads business goals from, noting most enterprises never wrote down whether a workload should trade cost for resilience. Alex Kargin, a senior data engineer, warned about the jump from recommendations to automation-ready fixes, since saving money can hurt resilience and improving performance can raise cost.
AWS acknowledged that AI-generated recommendations may contain errors or be incomplete and should be evaluated with oversight. The agent is available in three US regions, though workloads from any region can be onboarded, and it requires Business Support+ or higher. That last condition frames the product less as a tool for everyone and more as a tier upgrade.
