What the AI analyzes
- Alert signals — severity, affected services, golden signal type (errors, latency, saturation, traffic)
- Deployment changes — recent Kubernetes events and CI/CD deploys in the affected service’s cluster (via Change Intelligence)
- Service dependencies — services in the same environment that could be upstream/downstream causes
- Telemetry patterns — error rate spikes, latency increases, memory saturation from your OTel data
RCA output
Each RCA report contains:| Section | What it tells you |
|---|---|
| Summary | One paragraph: what happened, what was affected |
| Root cause | The most likely technical cause |
| Contributing factors | Secondary issues that made things worse or harder to detect |
| Impact | Who was affected and how severely |
| Recommendations | Specific actions to fix the root cause and prevent recurrence |
| Impact score | A 0–1 score reflecting severity and duration |
Triggering an RCA
Automatically from an alert
When a critical alert arrives via webhook, Obsy starts an RCA automatically. You’ll see it linked in the alert detail within seconds.From an incident
Open the incident detail and click Run RCA in the header. The analysis runs against the incident’s linked alerts, affected services, and timeline.Manually
Go to RCA in the sidebar and click New RCA. Select an alert or incident to analyze.Viewing RCA results
Go to RCA in the sidebar to see all analyses. Click any entry to open the full report. The report is also linked from:- The alert that triggered it
- The incident it’s associated with
- The postmortem (if one was created)