AI-Native SDLC · Edition 2026-09

The Maturity Matrix,
drawn to scale.

16 capabilities across 4 perspectives, each climbing five levels from Ad‑hoc to Autonomous. Open any cell to see what that level looks like in practice - criteria, evidence and guides included.

Capability ↓ · Maturity →

L1

Assisted

human writes · human checks

L2

Delegated

agent writes · human checks everything

L3

Systematic

agent writes · the system checks

L4

Governed

agent ships · evidence decides

where teams aim

L5

Self-improving

the system changes itself · humans audit

01DevelopmentHow developers work with AI day-to-day. From sidebar chat to fleet agents.→

Coding Agent Usage15 guides

L1Autocomplete and a chat window L2An agent in the IDE, rules in the repo L3CLI agents become the primary interface L4Unattended agents, inside a written boundary L5A fleet ships faster than you can read

Context Engineering15 guides

L1The agent sees one open file L2CLAUDE.md tells it the basics L3Context is served, not scavenged L4The org pushes context to the agent L5Context maintains itself

Code Review & Quality15 guides

L1Humans review everything, slowly L2AI suggests, humans still decide L3Lint is architecture; AI takes first pass L4Green merges on policy, never without an owner L5Human eyes only on Red

Testing Strategy15 guides

L1Tests by hand, flakes by habit L2Agents write tests, humans own the oracle L3Requirements are the oracle, not the code L4A red test means a real defect L5The suite heals itself

02Delivery ManagementHow we manage delivery in the age of agents. From human PR review to autonomous delivery pipeline.→

CI/CD Pipeline15 guides

L1CI runs; everyone waits L2Ten minutes, runners per team L3Five minutes, a pipeline per worktree L4Two minutes in an isolated microVM L5Feedback in seconds, capacity on demand

Merge & Deploy15 guides

L1Every merge has a human in it L2A queue does the rebasing L3Policy decides what merges, and when L4Green auto-merges straight to production L5A thousand merges a week, agent-driven

Metrics15 guides

L1DORA at best; AI still unmeasured L2Token spend is finally on a dashboard L3You know your cost per merged PR L4Every agent run ends in a known state L5Cost per feature, value per token

Governance & Compliance15 guides

L1Personal subscriptions, no policy L2A policy exists and spend has caps L3Every agent action leaves a trail L4Provenance is cryptographic, checks automated L5Compliance watches the regulators for you

03OrganizationHow organizations adapt to the age of agents. From "buy licenses" to "agent fleet management".→

AI Adoption Model15 guides

L1Licenses bought, little changes L2Pilot teams and a champion L3A platform team owns the tooling L4AI-first is the culture, not a memo L5The org is built around agent throughput

Knowledge Management15 guides

L1Knowledge lives in people's heads L2Docs and ADRs get written L3Documentation is infrastructure L4Context flows to agents automatically L5The knowledge base updates itself

Team Structure & Roles15 guides

L1Classic roles; seniors mop up AI code L2Champions and the first context engineers L3Harness and platform engineering are real jobs L4Developers manage fleets, not files L5Agentic engineers orchestrate; anyone contributes

Tech Debt & Modernization15 guides

L1Debt piles up, untouched L2Debt is at least triaged L3Agents pay debt down in the background L4Dead projects modernize for pennies L5Debt near zero, patched 24/7

04InfrastructureThe technical layer that enables (or blocks) agents. From shared Jenkins to ephemeral agent sandboxes.→

Agent Runtime & Sandboxing15 guides

L1Agents run on developer laptops L2Docker walls and scoped credentials L3Isolated devboxes, blind to production L4Hardware-isolated microVMs in ten seconds L5A fleet on its own compute

MCP & Tool Integration15 guides

L1Built-in tools and copy-paste L2A few MCP servers, wired by hand L3MCP is a managed platform L4Four hundred tools behind one door L5MCP is the nervous system

Build System15 guides

L1Full rebuilds, nothing cached L2Cached and parallel - that's it L3A dependency graph, remote and incremental L4Any change, under two minutes, per agent L5Builds too fast to notice

Observability & Feedback Loop15 guides

L1Logs and error alerts L2Structured logs, basic OpenTelemetry L3Agent cost and outcomes on dashboards L4Anomalies open tickets; agents investigate L5Production fixes itself through the loop

L1 Assisted · L2 Delegated · L3 Systematic · L4 Governed · L5 Self-improving

Not sure which cells are yours? Run the assessment and get your team placed on this chart in about an hour.

The Maturity Matrix,drawn to scale.

The Maturity Matrix,
drawn to scale.