Failures in Harness are not categorized today. There is no cross-team visibility when multiple teams hit the same root cause simultaneously. 29% of survey respondents (Feb 2026) rate root cause identification as POOR. Slack support data shows SonarQube and JFrog errors hitting multiple orgs with no cross-team pattern surfaced — every team re-discovers the same issue independently.

We need:

— Automatic failure categorization (infra, test, compile, config, external dependency)

— Customer-managed knowledge base: regex rules → known causes → remediation steps

— Failure trending and flaky test detection across builds

— MTTR tracking per failure category

— Cross-build correlation: surface when N pipelines fail on the same root cause

The knowledge base must be editable by the customer (DPE) without requiring Harness involvement.

Business value

Eliminates repeated ticket ping-pong. Target: MTTR from 2–4 hours to under 15 min for categorized failure types.