Budgets and model policies.
OneOrg lets admins choose which AI providers and models power each system capability, then use budget and usage views to keep model behavior governed at the organization level.
Set model policy
Admins configure model behavior from the AI provider, system model, budget, and policy settings areas. Provider settings define which external AI providers are available to the organization. System model assignments decide which provider and model OneOrg should use for specific capabilities, such as chat, agent generation, skill generation, widget generation, or slide generation.
This keeps model selection out of the hands of individual feature screens. A user can ask OneOrg to create an agent or generate a widget without choosing the underlying model each time. The organization decides the default model for that capability, and the product experience follows that policy.
Budgets and usage views help admins see where AI spend is coming from. They can review provider usage, compare model choices, and adjust assignments when a capability needs a different cost, latency, or quality profile. The configuration is meant to make model use deliberate rather than accidental.
Governance is the point. Model choice becomes an administrative policy, not a hidden side effect of whichever AI feature a user opened.
A good setup names the provider, assigns the model to a specific capability, then watches usage after real team activity. That gives admins a feedback loop: configure, observe, adjust.
Set model policy before broad AI rollout. Start with the capabilities users will touch most often, then assign models based on the actual tradeoff: quality for generation-heavy work, lower cost for routine drafts, and predictable latency for interactive flows. Review usage after teams have real activity instead of guessing the perfect setup up front. If spend spikes, do not only lower limits. Check whether the wrong capability is using an expensive model, whether a team is repeating failed prompts, or whether a workflow should move to a more constrained agent or skill. Budget controls work best when paired with clear model assignments and role limits.
Govern cost and quality
Use budgets and model policies when your organization wants consistent AI behavior across OneOrg. They are useful for standardizing providers, limiting expensive models to high-value capabilities, and reviewing which features are driving usage.
Do not treat them as personal model preferences. These settings govern organization capabilities. Usage outside OneOrg, or calls made directly against a provider account, sits outside OneOrg's budget and policy view.