davidfacer.com / aimaturitymodels.com / AI-Native Maturity Models / AI-Native SDLC / Whole-Model View
AI-Native SDLC — Whole-Model View
Thirteen dimensions, five maturity levels each. Select a cell to read it — click, or use the arrow keys. The detail is always open.
| Dimension | ANascent | BModeled | CContinuous | DIntegral | ETelemetric |
|---|---|---|---|---|---|
| D1Market discovery & definition | Ad hoc AI use, by preference, no shared method. | Consistent AI method exists, still one function's own. | AI discovery runs continuously now, still function-owned only. | Functions share one AI model, still hand-render it. | AI auto-renders each function's view from one model. |
| D2Buyer & user persona development | Ad hoc AI use, individual preference only, inconsistent. | Consistent AI synthesis exists, but personas ship static. | AI keeps personas live against real usage signal. | Functions consume shared AI persona model, still hand-adapting. | AI auto-generates each function's persona view directly. |
| D3Positioning & competitive intelligence | Ad hoc AI use, no shared method, inconsistent. | Consistent AI monitoring exists, one function's own cadence. | AI competitive monitoring is continuous now, not periodic. | Functions draw on shared AI model, still hand-rendering. | AI auto-generates each function's positioning view directly. |
| D4Requirements management | No AI role; informal tickets, no real structure. | No AI role; shared record, purely human traceability. | No AI role; shared human access to specs. | AI generates code and tests directly from spec. | AI generates both directions, spec to artifact, back. |
| D5Prototyping | No AI governance role; the boundary stays undefined. | AI-assisted prototyping exists; human decides case by case. | Human makes a documented carry-forward call, after building. | Same human criteria now applied while actively building. | Spec, prototype, and test built together; human call near-free. |
| D6Architecture governance | Tribal knowledge only; no AI involvement at all. | Humans write code; humans manually enforce it against docs. | Deterministic tooling, not AI judgment, catches common violations. | Agents apply the encoded rules during their own generation. | Agents propose rule changes; a human architect still approves. |
| D7Test-driven development | Humans write tests only after the code exists. | Humans write tests first, manually translated from the spec. | AI derives tests directly from the spec, repeatably. | AI generates tests as the spec itself is written. | Spec, test, and prototype generation fused, scaling with volume. |
| D8Code verification & review | Only human review exists; there is no automation. | Non-AI gates run first; humans still review everything uniformly. | Humans triage by risk; low-risk clears on automation alone. | An independent AI reviewer performs first-pass review, separately. | The AI reviewer's own accuracy is checked against outcomes. |
| D9Environment management | No environment separation exists; this is entirely manual. | Environments exist but are manually provisioned by a human. | Environments are defined as code; a human still triggers. | A human action triggers on-demand provisioning from templates. | Agents self-provision and tear down within human-set policy. |
| D10Release discipline | Releases are fully human-gated, on a fixed cadence. | Automated gates run; human approval and comms remain manual. | Deploys trigger automatically on passing gates; rollback stays exceptional. | Approval, compliance, and rollback are fully automated, policy-governed. | Live system signals govern expansion or rollback of releases. |
| D11Security & compliance | Agents act anonymously; nothing records what they did. | Every agent action is attributed and checked, not gated. | Agent floor is gated; each role's departures are read. | Earned agent refusals are mediated at runtime, non-bypassably. | Agent bounds evolve from telemetry on hits and silence. |
| D12Instrumentation & observability | Monitoring covers infrastructure only; no AI structuring exists. | AI generation produces some trace, but it goes unretained. | AI traces are retained systematically, diagnosing failure patterns. | AI's own role in runtime decisions is separately traced. | Build, runtime, and production telemetry are all equally visible. |
| D13Feedback loop velocity | No AI role; cycle time is simply unmeasured. | No AI role; humans measure cycle time retrospectively only. | Cycle time is tracked continuously; humans name the bottleneck. | Fixing the bottleneck is funded like ordinary feature work. | AI evidence and fixing compress response time to days. |
What D1 measures
The capability to identify, size, and validate market opportunities — including TAM/SAM/SOM, market trends, and opportunity framing — through a continuous, AI-assisted discovery loop that supplies a shared intelligence foundation for requirements, prioritization, and portfolio investment decisions.
Where it ends
Verify that automatic renderings accurately reflect the shared intelligence and that no consuming function is silently receiving stale or miscontextualized output. Audit D1–D3 convergence periodically to confirm that the layer is coherent, not merely synchronized.
⚠ = open review item
Currency basis: short-form cells pinned at df8735b; full model text pinned at de2668a. These are different bases, so this page can render content from more than one point in its sources' history.