Skip to content

AI_QA_MATURITY_MODEL.md

Latest commit

 

History

History
46 lines (34 loc) · 2.09 KB

AI_QA_MATURITY_MODEL.md

File metadata and controls

46 lines (34 loc) · 2.09 KB
Raw

AI-QA Maturity Model

This framework can be used as an assessment baseline for AI quality assurance maturity in delivery teams.

Levels

1. Ad hoc

  • AI quality is checked manually and inconsistently.
  • There are no stable requirements, gates, or reusable scenarios.
  • Release decisions depend on expert judgment rather than evidence.

2. Repeatable

  • Core prompts or user flows are tested repeatedly.
  • Basic regression checks exist for a few quality dimensions.
  • Results are visible, but not linked to formal requirements or risks.

3. Governed

  • Requirements, risks, and acceptance criteria are documented.
  • Scenarios are mapped to requirements and reused across runs.
  • Quality results can be explained to QA, product, and engineering stakeholders.

4. Release-Gated

  • Release thresholds exist for pass rate, critical requirements, and high-risk coverage.
  • Evidence supports GO, GO WITH RISKS, or NO-GO decisions.
  • Critical failures and uncovered requirements are explicitly visible.

5. Continuous Assurance

  • History and trend data are tracked across releases, models, or prompt changes.
  • Quality gaps feed a managed improvement backlog.
  • AI quality evidence becomes part of operational governance, audit, and vendor/model decisions.

Assessment Guidance

Use the maturity level as a communication aid, not as a substitute for the detailed report:

  • Ad hoc or Repeatable means the team is testing AI behavior, but governance is still weak.
  • Governed means the team can explain what is being tested and why.
  • Release-Gated means AI quality is materially influencing release decisions.
  • Continuous Assurance means quality data is being used over time, not only per run.

Typical Improvement Steps

  • Move from Ad hoc to Repeatable by defining a reusable scenario catalog.
  • Move from Repeatable to Governed by documenting requirements and linking them to scenarios.
  • Move from Governed to Release-Gated by adding explicit thresholds and release outcomes.
  • Move from Release-Gated to Continuous Assurance by tracking historical trends and operational follow-up.