1. IMR Help
  2.    
  3. AI in Xurrent IMR
  4.    
  5. AI Root Cause Agent

AI Root Cause Agent

Investigate the probable cause of an incident

The Root Cause Agent is Sera's investigation capability in Xurrent IMR. It reviews an incident and returns a hypothesis about what caused it, together with the evidence behind that hypothesis.

Each investigation returns a probable root cause, a confidence level, the evidence Sera drew on, and suggested mitigation steps.

Sera works from your own incident history. It reads the alert payload on the current incident, identifies the most similar incidents your team has already resolved, and reasons over the postmortems and chat threads attached to them.

Before you begin

Requirements

  • The incident must have an alert payload. Sera uses it as the starting point for finding comparable incidents.
  • To run investigations from a chat channel, the Slack integration must be configured for the team.

Important

  • Sera returns a hypothesis, not a diagnosis. Confirm it against live signals before applying any mitigation, particularly a rollback, failover, or restart.
  • Results improve with history. A new account with few resolved incidents has little for Sera to compare against, so early investigations return lower confidence and thinner evidence.
  • Postmortem quality directly affects investigation quality. Sera reads the postmortems your team writes, so incidents closed without one contribute nothing to future investigations.

Run an investigation

In Xurrent IMR

  1. Open the incident and go to the Details tab.
  2. In the Summary section, click Investigate incident.
  3. Review the hypothesis, confidence level, and evidence before acting on it.

In Slack

An investigation can be triggered directly from the incident's Slack channel, so responders do not need to leave the channel they are working in.

What an investigation returns

Output Description
Probable root cause Sera's hypothesis about what caused the incident
Confidence level How strongly the available evidence supports the hypothesis
Evidence The past incidents, postmortems, and chat threads the hypothesis is based on
Mitigation steps Actions that resolved the comparable incidents Sera identified

Root Cause Agent or Summarizer?

Task Use
Understand why the incident is happening Root Cause Agent
Describe what is happening, for someone who needs an update AI Incident Summarizer
Retrieve one specific field from an alert payload AI Querier
Document the incident after resolution AI Postmortem