Organization
Part VII. Organization: who owns observability
For: architects · team managersPrerequisites: Know the capabilities described in parts I to VI of the method.
In my view, this is where many implementations fail. GenAI observability straddles five functions, and without clear ownership nobody watches the right signals.
24. RACI
Section titled “24. RACI”R Responsible for execution, A Accountable, C Consulted, I Informed.
| Capability | Platform / SRE | ML / Data Science | Security / SOC | Product | Compliance / DPO |
|---|---|---|---|---|---|
| OTel instrumentation | A, R | C | I | I | I |
| Service SLOs | A, R | C | I | C | I |
| Quality SLOs | C | A, R | I | C | I |
| Evals and golden set | C | A, R | I | C | C |
| GenAI FinOps | A, R | C | I | C | I |
| MCP security | C | I | A, R | I | C |
| Audit and compliance | C | C | C | I | A, R |
| Quality on-call | R | A, R | I | C | I |
25. Quality on-call
Section titled “25. Quality on-call”A change from classical SRE: you need on-call coverage for quality regressions, not only for availability. A major difference: you do not restart a hallucination. The mitigation playbook is specific:
- roll back to the previous prompt or model version.
- tighten the guardrails (input and output filters).
- switch to a fallback model.
- disable the degraded feature rather than serve false quality.
These playbooks must exist before the incident, with an owner identified in the RACI.