Skip to main content
Incidents & Monitoring
4 min readUpdated 2026-02-15

Creating Incident Records

Step-by-step guide to creating incident records in Klarvo — capturing what happened, who was affected, severity assessment, containment actions, and resolution documentation.

Creating Incident Records

When an AI-related incident occurs, documenting it promptly and thoroughly is essential — both for regulatory compliance and for your own learning and improvement.

How to Create an Incident

  • Navigate to Incidents in the sidebar
  • Click Report Incident
  • Fill in the incident form (details below)
  • Click Save
  • Incident Form Fields

    Basic Information

    FieldGuidance
    TitleClear, descriptive: "Chatbot provided incorrect product safety information"
    Linked AI SystemSelect the AI system involved
    SeverityCritical / High / Medium / Low (see severity guide)
    StatusOpen / Investigating / Contained / Resolved / Closed
    Date OccurredWhen the incident happened
    Date DetectedWhen it was discovered
    Reported ByWho reported it

    Impact Assessment

    FieldGuidance
    Affected PartiesCustomers, employees, candidates, public, etc.
    Number AffectedEstimated number of impacted individuals
    Harm DescriptionWhat harm occurred or could have occurred
    Harm TypeSafety / Discrimination / Privacy / Financial / Reputational / Other

    Response Actions

    FieldGuidance
    Containment ActionsWhat was done immediately: "Disabled chatbot feature," "Reverted to manual process"
    Internal NotificationsWho was notified internally: Security, Legal, Compliance, Leadership
    External NotificationsProvider, market surveillance authority (if serious incident)
    Suspension StatusWas the system suspended? Fully / Partially / No

    Resolution (completed after investigation)

    FieldGuidance
    Root CauseWhat caused the incident
    Resolution ActionsWhat was done to fix it
    Preventive MeasuresWhat will prevent recurrence
    Date ResolvedWhen the incident was resolved
    Lessons LearnedKey takeaways

    Attaching Evidence

    For each incident, attach supporting evidence:

  • Screenshots of the problematic AI output
  • Logs showing the inputs and outputs involved
  • Communication records (notifications sent)
  • Postmortem reports
  • Corrective action documentation
  • After Creating

    Once an incident is logged:

  • It appears in the AI system's incident history
  • Critical/High severity incidents trigger a reassessment flag
  • Tasks are auto-created for investigation and resolution follow-up
  • The incident count is reflected on the dashboard
  • Best Practices

    ⏱️ Log immediately: Capture facts while they're fresh — you can update details later
    📸 Attach evidence: Screenshots and logs are crucial for investigation
    🔍 Be specific on harm: "Incorrect information provided" is less useful than "Customer received incorrect safety dosage guidance for medication"
    📋 Follow through: Don't close incidents without documenting resolution and preventive measures