Agentic DevOps Maturity Model¶
david-sanchez's four-level ladder for agent-readiness.
| Level | Foundations | Agent Adoption | Pipeline | Governance |
|---|---|---|---|---|
| 1. Reactive | Manual deploys, inconsistent tests, no IaC | Ad hoc AI assistant use by individuals | Basic CI | None |
| 2. Foundation | Automated CI/CD, IaC, security scanning, branch protection | IDE-level AI team-wide + shared instruction files | Standard PR verification + human review | Basic AI-tool policies; no formal agent gov |
| 3. Structured | Rich skill profiles, spec-driven dev, tests at scale | Custom agents; PR attribution metadata | Agent-specific verification + scope + provenance | Formal governance, auditability, delegation chains |
| 4. Optimized | Living specs; continuous compliance; platform engineering | Agent teams across lifecycle; remediation loops | Adaptive depth; pipeline-as-specification; attestation | Continuous gov; agent-native observability |
Sanchez's leverage advice (verbatim)¶
- Level 1 → 2: "The highest-leverage move for teams at Level 1 is not to adopt agents; it is to invest in the DevOps foundations that make agents effective."
- Level 2 → 3: spec-driven dev + repo skill profiles.
- Level 3 → 4: pipeline transformation + formal governance.
"Most organizations today are between Levels 1 and 2."
Why this matters¶
Turns agents-scale-not-fix from slogan into diagnostic. Gives engineering leaders a self-assessment tool with prescriptive next-step investments rather than a vague "adopt AI" mandate.