Policies
Put enforceable guardrails in front of AI spend.
Policies help teams control which models can be used, where budget can be spent and which requests require restriction, fallback or review.
What policies can control
Budget limits
Set company, team, project or workflow-level spend boundaries.
Provider access
Allow or restrict providers based on governance, data handling or commercial requirements.
Model tiers
Reserve premium models for complex, critical or approved workloads.
Fallback behaviour
Define how requests should behave when a model, provider or policy path is unavailable.
Policy lifecycle
Define Simulate Apply Monitor Refine
Implementation guidance
Start with broad production-safe defaults: server-side keys only, budget caps, provider allow-lists and a clear fallback policy. Add tighter rules once audit data shows real usage patterns.