AI-Assisted Incident Investigation
Standardised Telemetry
Summary
AI tools and human responders use the same telemetry, without separate instrumentation or exports for AI-assisted incident investigation.
Standards
- A service MUST NOT maintain bespoke instrumentation or telemetry exports designed specifically to support AI-assisted incident investigation.
std-ai-standardised-telemetry-01 - An AI tool used for incident investigation MUST draw on telemetry from the same observability platform used by human responders.
std-ai-standardised-telemetry-02
Related Standards
- Telemetry Instrumentation: Instrumentation Standards
- Distributed Tracing: Trace Context
- Observability Platform Integration: Telemetry Ingestion
Implements These Principles
Human Verification & Approval
Summary
AI-assisted incident remediation is applied or approved by a human responder.
Standards
- Before using an AI-generated diagnosis to guide high-impact remediation, a responder SHOULD review the logs, metrics, traces or change records cited by the tool.
std-ai-human-verification-approval-01 - A remediation or configuration change proposed by an AI tool MUST be applied by a responder or have a responder's approval before it is applied.
std-ai-human-verification-approval-02
Related Standards
Implements These Principles
- Automation Supports Human Judgement
- People Remain Responsible
- Independent Checks of AI Work and Evidence
Incident Traceability
Summary
Teams record AI-generated findings and advice used in incident response and should review and address recurring inaccuracies.
Standards
- AI-generated findings or advice used in incident response MUST be recorded in the incident record, directly or by linking to retained findings.
std-ai-incident-traceability-01 - Engineers SHOULD review recurring errors in AI-generated findings and provide corrective feedback or update the tool's context or configuration.
std-ai-incident-traceability-02