Build an incident triage agent this afternoon

AWS published a tutorial on June 9 for building an incident triage agent using Amazon Quick, New Relic, and Asana. The agent investigates alerts, drafts root-cause analysis briefs, and files tracked tasks, reducing evidence-gathering time in internal tests. The tutorial targets engineering teams with enterprise SaaS subscriptions.

On 9 June, AWS published a step-by-step tutorial https://aws.amazon.com/blogs/machine-learning/build-an-agentic-incident-triage-assistant-with-amazon-quick-and-new-relic/ for an incident triage agent built on Amazon Quick — AWS’s enterprise chat-agent builder — that pulls data from New Relic an observability platform and files a tracked Asana task a project-management handoff . The agent investigates alerts, drafts a root-cause analysis RCA brief with evidence links back into New Relic, and asks for confirmation before creating the task — all from a single on-call engineer’s prompt. The post targets engineering teams running on-call rotations. AWS says the agent reduced the evidence-gathering phase of triage in internal testing at New Relic, giving faster resolution and a consistent investigation standard across shifts. How the agent works The agent is a custom chat agent in Amazon Quick with two built-in connectors: New Relic — a connector added to Quick in May — and Asana. From a prompt like Checkout is slow and we are seeing server errors on checkout-service in production, check the last 24 hours, generate RCA brief , the agent: - Calls five New Relic reasoning tools and decides which to run based on the prompt. - Assembles an RCA brief in a fixed format — summary, blast radius, likely trigger, key evidence with links, and three recommended next actions. - Asks for confirmation before filing an Asana task tagged sre-triage, incident in a project called SRE Incident Triage . The New Relic tools do the actual investigation; the agent routes between them. Prerequisites This is enterprise plumbing. To run the tutorial as-written you need: - An Amazon Quick Professional subscription with Author-level permissions — so a paid seat, not a free tier. - A New Relic account you can authenticate the connector against. - An Asana workspace with a project named SRE Incident Triage , plus admin access to the Asana developer console to register an integration. If your team doesn’t already pay for Quick Professional, the cost-of-entry alone puts this outside the reach of a sole trader or a typical 8-person services firm. What to do with this If you’re an engineering team already paying for Amazon Quick Professional and New Relic, this is a half-afternoon prototype — the steps are concrete, and you end with a working agent you can put in front of your on-call rotation. Three things to weigh before you build it: The stack cost. Quick Professional plus New Relic plus Asana plus the connector plumbing is enterprise SaaS. A sole trader or 8-person services firm won’t be touching this; an internal champion at a 50–250-person SME only justifies it if on-call triage is already a measurable pain. Trust boundaries. AWS recommends a dedicated, read-only New Relic service account and a tightly scoped Asana service account. Don’t bolt personal admin credentials onto a chat agent that can create tasks in a shared project. Your handoff tool. If your team runs Jira, Linear or GitHub Issues rather than Asana, the final step won’t translate directly. Quick has other action connectors, but the tutorial as written ends in Asana. New Relic is also building its own assistant in the same vein — SRE Agent , currently in preview inside the New Relic platform — and a customer described it doing in seconds what their team had taken nearly a year to figure out manually. “It took almost a year for us to figure out a very complex performance problem. The SRE Agent just picked it up by itself. I mean, we have pretty good instrumentation around everything now, but the ability for it to discern the problem and give us a path to the solution… I thought it was incredible.” — Site Reliability Engineer at a large healthcare enterprise, quoted in New Relic’s SRE Agent launch post If you’re not on the AWS stack, the underlying pattern — a chat agent that calls an observability tool and writes a tracked handoff task — is the part worth borrowing. Our coverage of CrewAI Flows /articles/crewai-flows-event-driven-orchestration/ and LangChain’s Deep Agents /articles/langchain-deep-agents-pattern/ sits in the same vein, and our governance piece /articles/microsoft-agent-governance-toolkit/ covers the guardrails you’ll want in place before any agent creates tickets on your team’s behalf. Sources & quotes Every quotation in this article is verbatim from a named source — click any 1 to see where it came from. It's part of how we keep an AI-run newsroom honest. How we verify → /blog/how-we-keep-an-ai-newsroom-honest/