AI Product Maturity Checklist
Assess your AI product against AIPCH01–AIPCH20. Your answers are saved locally (browser only).
Scoring uses percentage normalization since weights total 92.
🤖 AI Product Maturity Checklist
0.0% — AI Asset (Non-Productized)Mark each characteristic as Yes, Partial (50%), or No. Add optional evidence links/notes.
| # | AIPCH | Characteristic | Checklist Item | Description | Weight | Status | Evidence (URL or note) | Points |
|---|---|---|---|---|---|---|---|---|
| 1 | AIPCH01 | Domain-Owned | Named AI Product Owner (AIPRO) Assigned | Product is owned by a specific business domain or functional area with a named accountable AIPRO. | 5% | 0 | ||
| 2 | AIPCH02 | Deployable | End-to-End Designable, Developable, Deployable by Domain | Domain team can design, train, and deploy the AI Product independently using self-service tooling. | 5% | 0 | ||
| 3 | AIPCH03 | Declarative | Defined via Declarative AI Specification | Model architecture, configuration, policies, and interfaces are expressed through structured specs (AIPDS YAML/JSON). | 4% | 0 | ||
| 4 | AIPCH04 | Discoverable | Registered in AI Marketplace or Catalog | Product is discoverable through metadata-rich listings including capability, domain, and trust score. | 4% | 0 | ||
| 5 | AIPCH05 | Self-Describing | Rich Metadata Embedded (Model Card / Datasheet) | Includes description, version, training summary, input/output schema, and usage notes. | 4% | 0 | ||
| 6 | AIPCH06 | Trustworthy (via Trust Signals) | Emits Trust & Quality Metrics | Model publishes trust metrics such as accuracy, drift, compliance, and ethical posture. | 5% | 0 | ||
| 7 | AIPCH07 | Reusable | Reused Across Use Cases or Domains | AI Product is reused or extended across multiple workflows, applications, or teams. | 4% | 0 | ||
| 8 | AIPCH08 | SLA–SLO Backed & Observable | Performance Monitored and Measured | Latency, inference accuracy, uptime, and quality metrics are defined and continuously observable. | 5% | 0 | ||
| 9 | AIPCH09 | Compliant by Design | Policy-as-Code Enforcement for AI Ethics & Regulation | AI Product integrates regulatory and ethical compliance checks (GDPR, fairness, bias mitigation). | 5% | 0 | ||
| 10 | AIPCH10 | Addressable | Provides Well-Defined Interfaces | Exposed via APIs, SDKs, or agent interfaces with standard documentation and access control. | 4% | 0 | ||
| 11 | AIPCH11 | Semantically Aligned | Aligned to Enterprise and Domain Ontologies | AI Product uses canonical vocabularies and semantics for input/output data. | 4% | 0 | ||
| 12 | AIPCH12 | Consumption-Driven Intent | Built for Clear Business Purpose or Consumer Context | Product’s design is guided by explicit business or operational use cases. | 4% | 0 | ||
| 13 | AIPCH13 | Testable & Versioned | Includes Evaluation Dataset and Version Control | Product maintains reproducible tests, benchmark datasets, and semantic versioning. | 5% | 0 | ||
| 14 | AIPCH14 | Economically Accountable | Tracks Cost, Usage, and ROI Metrics | Inference, training, and infrastructure costs are tracked, attributed, and visible to stakeholders. | 5% | 0 | ||
| 15 | AIPCH15 | Explainable & Transparent | Model Decisions Are Interpretable and Documented | Includes explanation interfaces, LIME/SHAP-style outputs, and global feature insights. | 5% | 0 | ||
| 16 | AIPCH16 | Bias-Controlled & Fairness-Measured | Bias Metrics Monitored and Actively Mitigated | Includes fairness evaluation, bias dashboards, and mitigation workflows. | 5% | 0 | ||
| 17 | AIPCH17 | Continually Learnable (Retraining Ready) | Supports Continuous Improvement Loops | Product supports retraining pipelines with drift detection and feedback integration. | 5% | 0 | ||
| 18 | AIPCH18 | Safe & Policy-Bound Usage | Prohibited Use Policy & Safety Controls Implemented | Includes explicit boundaries for use, misuse detection, and ethical filters. | 5% | 0 | ||
| 19 | AIPCH19 | Interoperable & Composable | Can Be Orchestrated with Other AI or Data Products | Compatible with APIs, orchestration frameworks, and hybrid agent systems. | 4% | 0 | ||
| 20 | AIPCH20 | Human-Centered Oversight | Supports Human-in-the-Loop Review and Override | Includes escalation, appeal, and override mechanisms for critical decisions. | 5% | 0 | ||
| TOTAL | 0.0 | |||||||
🎯 Prioritized Recommendations
High Priority — Not Met
- AIPCH01 (5%): Assign an accountable AIPRO and publish ownership, escalation, and lifecycle metadata in the registry.
- AIPCH02 (5%): Provide self-service pipelines for design→train→deploy; ensure environment parity and rollback paths.
- AIPCH06 (5%): Instrument accuracy/quality and compliance signals; expose trust score and evaluation history.
- AIPCH08 (5%): Define SLIs/SLOs; create dashboards; set alerts; add runbooks for incidents.
- AIPCH09 (5%): Implement privacy/residency/consent checks; fairness gates; audit logging; retention policies.
- AIPCH13 (5%): Version datasets/models; run regression/eval gates in CI; publish changelog and deprecation policy.
- AIPCH14 (5%): Integrate FinOps; report per-1k tokens/inference cost; enable showback/chargeback; budget alerts.
- AIPCH15 (5%): Provide local/global explanations; document decision factors; expose explanation endpoints/UI.
- AIPCH16 (5%): Run fairness audits; monitor subgroup metrics; apply mitigation; publish bias reports.
- AIPCH17 (5%): Implement drift detection; capture feedback loops; automate retraining/re-evaluation pipelines.
- AIPCH18 (5%): Publish prohibited-use policy; add safety filters/red-team mitigations; log and review overrides.
- AIPCH20 (5%): Provide override/escalation workflows; record interventions; include appeals process and audit trails.
- AIPCH03 (4%): Adopt AIPDS; validate spec in CI; generate infra/policies from spec; remove ad-hoc manual steps.
- AIPCH04 (4%): Onboard to registry; include task type, modalities, owners, risk class, trust score, and quickstart examples.
- AIPCH05 (4%): Publish comprehensive Model Card/Datasheets; link IO schema, training summary, and limitations.
- AIPCH07 (4%): Ship SDKs and adapters; document integration patterns; capture and showcase reuse examples.
- AIPCH10 (4%): Publish OpenAPI/gRPC; provide SDKs; document auth, quotas, and example notebooks.
- AIPCH11 (4%): Map IO schemas to enterprise/domain ontology; use persistent identifiers and lineage links.
- AIPCH12 (4%): Define primary use-cases and KPIs; capture acceptance criteria; provide task-specific examples.
- AIPCH19 (4%): Ensure composable interfaces; agent/flow orchestration compatibility; schema adapters.
ℹ️ Scoring Guidelines
- Yes = full weight
- Partial = 50% of weight
- No = 0%
≥ 80%: Production-Grade AI Product · 50–79%: Evolving / MVP AI Product · < 50%: AI Asset (Non-Productized)
🔐 Governance Tips
- Link to AIPDS spec, Model Cards, eval reports, and safety/policy repos.
- Export JSON for audit trails; attach to PRs or registry entries.
- Use prioritized recommendations to plan next-quarter hardening work.