Add automated, pre‑validated remediation after AWS DevOps Agent investigations

This reduces the time your on‑call team spends turning a diagnosis into a fix by converting AWS DevOps Agent investigation summaries into pre‑validated remediation steps you can run automatically or approve.

What actually changed

The AWS post shows a way to close the loop from diagnosis to action. AWS DevOps Agent continues to produce root‑cause analysis (RCA) and recommended actions; the new pattern layers AWS Lambda Durable Functions, Amazon EventBridge and Amazon Bedrock to take those investigation summaries and drive an automated remediation workflow. The post demonstrates using the three services together so the system can convert investigation output into artefacts intended to support automated remediation rather than leaving the final step to a human on call.

Who it affects

This affects teams that already run AWS DevOps Agent in an "observe‑and‑report" mode and want to reduce the manual work between investigation and remediation. If your on‑call rotation spends time diagnosing across logs, metrics and topology and then executing fixes, this pattern shifts the repetitive translation work into an orchestrated workflow that can be previewed, pre‑validated and executed under your controls.

What it replaces or costs

What this pattern replaces: the manual handoff where an engineer reads an investigation, crafts a fix and applies it. The blog frames the workflow as a complement to the DevOps Agent that completes the issue‑resolution step for you.

What it costs: the announcement does not state pricing. The components mentioned — AWS Lambda Durable Functions, Amazon EventBridge and Amazon Bedrock — each carry their own billing models, and you will also need to account for any compute, invocation and Bedrock model use in your cost estimates.

Practical steps you can do this week

These steps follow the architecture described in the post and keep the agent in observe‑and‑report mode while you build a safe automation path.

1. Capture investigation output

• Confirm AWS DevOps Agent is producing investigation summaries and where they are emitted (logs, events or a service endpoint). Keep it in observe‑and‑report mode while you prototype.

1. Route summaries into a durable orchestrator

• Create an Amazon EventBridge rule or event path that receives the investigation summaries and sends them to a Lambda Durable Functions orchestrator. The post uses Lambda Durable Functions as the orchestrator to manage multi‑step remediation flows.

1. Generate remediation artefacts with Bedrock

• From the orchestrator, call Amazon Bedrock to turn the investigation summary into candidate remediation steps or a remediation plan. Treat Bedrock output as a draft that requires automated validation or human review before execution.

1. Add pre‑validation and approval gates

• Implement automated pre‑validation tests in the Durable Functions workflow (for example: dry‑run checks, policy scans, or canary simulation). Add a manual approval step if you want human control before applying changes.

1. Execute or escalate

• If pre‑validation passes and any required approvals are given, have the orchestrator invoke the remediation playbooks or runbooks (Lambda, Systems Manager, or your existing tooling). If validation fails, route a clear incident update back to your on‑call channel.

What we don't know

• The exact event format or API surface AWS DevOps Agent exposes for downstream routing.
• How the post implements pre‑validation and what tests it runs.
• Whether the blog's example executes changes automatically or leaves execution behind an approval.
• Any performance or cost numbers for Bedrock, Durable Functions or EventBridge in this pattern.

What to do next

1. Pick a recurring incident type your team handles and keep AWS DevOps Agent in observe‑and‑report mode while you capture several investigation summaries for testing.
2. Build a minimal prototype: an EventBridge rule → Lambda Durable Functions orchestrator → call to Bedrock producing a candidate remediation, and wire in one automated pre‑validation step.
3. Run the prototype in a staging account, review failures and add a manual approval gate before enabling any automatic execution.