Skip to content

Roles and responsibilities

ORF is a shared operating model. Engineering produces the signals, platform teams make them dependable, operators use them, and accountable owners decide what evidence is necessary.

Core responsibilities

Role Accountable for Typical evidence
Product owner Intended use, users, service outcomes, impact, risk acceptance System profile, scope approval, prioritised gaps
Technical owner Architecture, instrumentation, schema, integration, technical acceptance Telemetry design, trace samples, tests
AI or application engineer Model, prompt, retrieval, agent, and tool events Interaction and tool records, evaluation results
Platform or SRE Collection reliability, storage, dashboards, alerts, retention Pipeline health, SLOs, alert tests
Service operations Triage, escalation, handover, incident response Runbooks, exercise and incident records
Security and privacy Threat signals, access, integrity, classification, minimisation Access reviews, privacy tests, audit-trail evidence
Risk or governance Risk tier, permitted waivers, control evidence, periodic review Approved register, waiver decisions, review minutes
Data or evaluation lead Baselines, datasets, sampling, labels, metric interpretation Evaluation protocol, dataset card, metric reports

Lightweight RACI

A = accountable, R = responsible, C = consulted, I = informed.

Activity Product Technical Platform/operations Security/privacy Risk/governance
Define intended use and boundary A/R C C C C
Determine risk tier A C I C R
Select requirements and delivery plan A R C C C
Decide permitted waivers C R C C A
Design event and trace model C A/R R C I
Approve sensitive-content handling I R C A/R C
Implement and test I A/R R C I
Operate alerts and runbooks I C A/R R I
Review operating evidence A R C C C
Decide residual-risk acceptance C C I C A/R

One person may hold several roles

Small teams do not need separate people for every row. They do need explicit ownership and an independent challenge for consequential risk, waivers, and go-live decisions.

Governance forums

Use existing forums where possible:

  • service operations for weekly reliability and incident review;
  • change advisory or release review for model, prompt, policy, data, and tool changes;
  • security/privacy review for access, leakage, credential, and audit-trail concerns; and
  • AI governance or risk oversight for scope, waivers, residual risk, and periodic assurance.

Done when: every selected ORF requirement, dashboard, alert, runbook, waiver, and review has one named accountable owner and a backup or escalation route.

Next: Understand your system