Operating Model
Observability is not owned by one team. It is a shared responsibility model with clear boundaries, aligned to G.A.I.N AIOM plane thinking.
RACI summary
| Activity | Product | Engineering | Platform | SRE | Architecture |
|---|---|---|---|---|---|
| Define business KPIs and journeys | A/R | C | I | C | C |
| Map KPI → service entry | C | R | I | C | A |
| Service instrumentation | I | R | C | C | A |
| Infra / K8s / DB telemetry | I | C | R | C | A |
| Correlation standards | C | C | R | C | A |
| Alert registry and paging | C | C | C | R/A | C |
| Dashboards (canonical) | C | C | R | R | A |
| Maturity roadmap | C | C | C | C | A/R |
| Audit / replay consumers | C | C | R | C | A |
R = responsible · A = accountable · C = consulted · I = informed
Team charters (one paragraph each)
Product
Owns what success looks like for customer and regulatory journeys. Defines outcome KPIs, business SLOs, and journey steps. Does not own Prometheus or log pipelines.
Engineering
Owns service instrumentation: golden signals, traces, dependency accuracy, and feature-level SLIs. Ships code that emits structured logs and propagates correlation IDs.
Platform
Owns observability infrastructure: collectors, stores, retention tiers, enrichment, cost attribution, and self-service onboarding templates.
SRE / reliability
Owns alerting, incident response, and error budgets. Maintains alert registry, runbooks, and drill programs. Escalates when business SLOs burn.
Architecture
Owns standards and the correlation model: governance rules, maturity targets, tool boundaries, and cross-domain graph design. Resolves layer-mixing debates (strategy vs system vs software).
Cadence
| Meeting | Frequency | Output |
|---|---|---|
| SLO review | Monthly | Error budget decisions |
| Journey / KPI review | Quarterly | Updated business maps |
| Maturity assessment | Semi-annual | Roadmap refresh |
| Graph drill | Quarterly | E2E trace evidence |
AI-specific overlay
When agents or RAG are in scope, add:
| Role | Added responsibility |
|---|---|
| AI platform | Request-path spans, eval sampling hooks |
| Security / compliance | Audit tier access, redaction policy |
| Model risk (regulated) | Quality metrics as control evidence |
Bridge: G.A.I.N Observability · Eval Engineering playbooks.
Release gate
- RACI published for observability workstreams
- Named owners for tier-1 journeys and services
- Monthly SLO review on calendar
- Architecture sign-off on correlation standard