Institute/Workflows
Every workflow below is a documented procedure with named artifacts and observable signals. Each one maps to a dimension of the AI-QA Maturity Model. These are the procedures the readiness audit scores against, and the procedures the retainer runs on your code.
Eval-suite design, golden-set curation, and baseline tracking. The discipline that turns "did you try it?" into "did the suite pass and did the baseline hold?"
Read the workflow → 02 · WorkflowPre-deploy gate specification, threshold enforcement, and the documented override policy. The procedure that decides what's allowed to ship.
Read the workflow → 03 · WorkflowProduction sampling, online evals, and the paging integration. The procedure that catches model degradation before users do.
Read the workflow → 04 · WorkflowTagged incident postmortems and the named-failure-mode catalog. The procedure that turns last quarter's pain into next quarter's eval coverage.
Read the workflow → 05 · WorkflowIn-product user reports, the triage queue, and the time-to-coverage SLO. The procedure that closes the loop from production failure to eval suite.
Read the workflow → 06 · WorkflowWritten refuse list, system-prompt enforcement, refusal-correctness evals, and the dated compliance review. The procedure that names what the AI must never do.
Read the workflow →In active development. Each workflow page below is a thin first cut: the name, the maturity dimension it maps to, and the procedure shape. The full procedural documentation — including the audit-evidence checklist per workflow — lands in Phase 2.
The five-minute readiness self-assessment scores you across all six. You'll get your weakest workflow flagged with a concrete picture of what the next level looks like.
Take the assessment →