AI Agent Orchestration Platform  ·  View 22 of 32  ·  4 · Runtime

Policy and Budget Enforcement

Five policy scopes against five decision points: exactly what is evaluated, and where.

Editable source SVG draw.io All views
At admission Before model call Before tool call Before response At completion Organisation Allowed regions Approved providers Banned tool classes Content restrictions Retention policy Project Monthly budget Model allow-list Tool allow-list Capture opt-in Cost attribution Agent Run concurrency Token ceiling Granted scopes Output schema Per-agent spend Workflow Max duration Recursion depth Approval steps Compensation rule Run cost record Data classification Residency pinning Route constraint Egress denial Redaction rule Deletion trigger Policy and Budget Enforcement — Scope against Decision Point Security / platform Decision point Data store Application we own One Rego bundle, five evaluation points. Decisions are local to the worker, so a control-plane outage cannot fail them open. v 1.0 · owner Governance · date 2026-08

Decisions

  • Policy is authored centrally and evaluated locally: the control plane signs a Rego bundle, a sidecar in every worker evaluates it
  • Local evaluation keeps decisions sub-millisecond and means a control-plane outage cannot cause policies to fail open
  • Scopes compose most-restrictive-wins; a project cannot widen what the organisation forbade

Governed by policy

  • Which models, which tools, which data, which actions need approval, maximum spend, maximum duration, maximum tokens, allowed regions and providers, retention, and content restrictions
  • Policies are versioned, auditable and evaluated at runtime, at organisation, project, agent and workflow level
  • Azure Policy governs the infrastructure; this engine governs agent behaviour, and the two are deliberately not the same system

Risks

  • A signed bundle can go stale if a worker cannot reach the publisher; bundles carry a maximum age and a worker refuses new work past it
  • Rego is a specialist skill; policy authoring is owned by a small central team with a reviewed template library
◀  Safety Guardrails All views Deployment Topology  ▶
Solution Architecture v1.0 · Data & AI Global Practice · 2026-08 · 32 views · Microsoft Azure22-policy-and-budget