AWS DevOps Agent gained two significant new capabilities in preview this month: Release Readiness Review and Autonomous Release Testing. Where the agent previously focused on operations after deployment, these features push it earlier into the delivery pipeline, evaluating whether a code change is actually ready for production and generating tests tailored to that specific change before it merges. Both are available in preview at no additional cost, currently limited to the US East (N. Virginia) region.
What Release Readiness Review checks
Rather than running a static linter or a fixed checklist, Release Readiness Review builds a knowledge graph of an organisation’s connected repositories to understand how services actually depend on one another, then evaluates a proposed change against production requirements, cross-repository dependencies, and internal engineering standards. Those standards can be defined in plain natural language rather than a dedicated policy-as-code framework, covering security, compliance, networking, observability, and operational rules specific to the organisation, alongside AWS Well-Architected best practices. The goal is catching a change that looks fine in isolation but would break something two services away, the kind of failure static analysis alone typically misses.
What Autonomous Release Testing adds
Autonomous Release Testing generates and runs test plans built specifically for each code change rather than executing a fixed, manually maintained suite. The agent reasons about what the change actually does and constructs tests around functional correctness, behavioural regressions, and integration scenarios a static test plan was never written to anticipate. Combined with Release Readiness Review, the pitch is a pipeline that evaluates both whether a change meets organisational standards and whether it actually works, before a human reviewer even opens the pull request.
Why AWS is building this now
AWS is candid about the problem driving this: AI coding assistants have dramatically increased how much code and how many pull requests get generated, while the review, compliance, and release validation steps around that code have stayed manual and have become the actual bottleneck. Writing code got faster; shipping it safely did not. That is not an AWS-specific problem, it is the shape of software delivery across the industry right now, and it explains why every major cloud provider is racing to put an agent somewhere in the pipeline between commit and production.
What this means for engineering teams
For teams evaluating this, the interesting question is not whether an AI agent can flag issues, most scanning tools already do that to some degree. It is whether natural-language engineering standards enforced by an autonomous agent hold up under audit, particularly for European organisations subject to NIS2, DORA, or sector-specific change management requirements that expect a documented, repeatable rationale for what shipped and why. A preview is exactly the moment to test that against your own standards before it becomes the default way changes get approved.
If your team wants help evaluating AWS DevOps Agent’s new release management capabilities, or needs a governance model for where AI-driven release gating fits into your existing compliance and change management processes, contact Excello Digital. We help European engineering teams adopt AI-native DevOps tooling without losing the audit trail regulators expect.