AI and Intelligent Automation
What the domain tests
AI and Intelligent Automation covers anomaly detection, alert correlation, the New Relic AI assistant, and AI Monitoring for LLM applications. The exam tests judgment: when automation reduces noise, and when a model or correlation is not trustworthy.
Core rules
- Baseline alert conditions learn a signal's normal range. They need sufficient history and stable seasonality; low-volume or spiky signals produce false positives.
- Incident Intelligence correlates related alerts into issues using entity relationships and timing, reducing alert fatigue. Alert quality reporting shows noisy conditions.
- AI Monitoring instruments LLM calls to capture model, token counts, cost, latency, errors, and feedback. It requires SDK or integration instrumentation; it is not automatic for every app.
- The New Relic AI assistant generates NRQL, explanations, and troubleshooting suggestions. Treat output as a draft: verify query scope, time window, and permissions.
- Automate with alert workflows, NerdGraph, and Terraform only after validating conditions and ownership.
Worked example
A checkout failure produces 30 alerts across four services. Incident Intelligence groups them into one issue with a probable originating service. Reason: shared entity relationships and timing indicate one cause; keep the issue, tune the noisy downstream conditions, and add a baseline condition on error rate only after checking volume.
Traps
Trusting generated NRQL without checking filters; baseline alerts on low-traffic endpoints; correlating unrelated alerts by time alone; expecting LLM cost metrics without instrumentation; automating remediation before validation.