User reports, triage queue, time-to-coverage SLO.
A failure that reaches production is half-broken QA. The remaining work is the feedback loop — getting that failure into the eval suite before it happens again. Without the loop, every L-level incident is forever; with it, last week's hallucination becomes this week's failing test case becomes next week's passing baseline. This workflow is what makes the maturity model an actual ladder instead of a static photograph.
The Feedback Loops workflow is the procedure for capturing user-reported failures in-product, routing them into a triage queue with a named owner, turning triaged failures into eval cases on a measured cadence, and tracking time-from-report-to-eval-coverage as a tracked SLO.
The five questions on the readiness self-assessment that score this dimension are the five rungs of the procedure above. Yes on a question means the artifact named in that step exists on disk in your repo today.
This page is a thin first cut. Full procedural documentation — including reference DeepEval suite scaffolds, golden-set curation rubrics, and the audit-evidence checklist — lands in Phase 2 of the Institute build-out.
The free readiness self-assessment scores the Feedback-Loop workflow as one of six dimensions. Five minutes. Your weakest workflow is the one most worth fixing first.
Take the assessment →