AWS DevOps Agent: AI‑Driven Incident Investigation

AWS DevOps Agent is an agentic AI system that autonomously investigates incidents by querying CloudWatch logs, New Relic telemetry, and recent deploy events, then posts a ranked investigation summary to Slack. The agent is GA but only available in us‑east‑1, us‑west‑2, ap‑southeast‑2, ap‑northeast‑1, eu‑central‑1, and eu‑west‑1. It requires a Full Platform User API key for the New Relic Telemetry Source and a Capability Webhook to trigger investigations automatically. The payload template must convert the action field from uppercase to lowercase and transform the service field from an array to a string. A custom skill can set the investigation priority order, ensuring the agent checks external dependencies first, then recent deployments, and finally internal metrics. The agent also correlates deploy events via a GitHub integration and posts findings to a dedicated Slack channel. The setup involves creating an Agent Space, wiring up New Relic, GitHub, and Slack, and uploading a custom skill. The agent runs in a separate web app accessed via an IAM‑authenticated link, while the console is used only for admin and setup.

During the PoC, the author discovered that us‑west‑1 is not a supported region, but creating the Agent Space in us‑west‑2 still gives cross‑region discovery. The agent automatically mapped over 4,000 entity relationships, including services, ECS clusters, ALBs, and Lambda functions. The author spent significant time debugging the payload template, correcting the action field casing and the service field type. The agent’s custom skill allows prioritizing external dependency checks, which proved effective in identifying third‑party outages.

The result is a first‑pass investigation that saves engineers the time spent gathering context, allowing them to arrive with actionable insights rather than starting from scratch.