Skip to content

Administrative model governance

Intric offers centralized governance of model selection to optimize cost efficiency. As an administrator, you can configure the system so that certain high-performing — and therefore more expensive — models require a specific approval before they can be used.

The process works as follows:

  1. Request — When a user wants to use a restricted model in an assistant, they must provide a justification for the use case.
  2. Review — A notification is sent to the central administrator (technical expert), who assesses whether the model choice is justified given the assistant’s purpose.
  3. Approval — Once approved, the model is activated for that specific use case.

This approach ensures organizational control and a clear validation process, because each assistant and use case is assessed on its own merits and potential for creating value. It reduces the risk of an individual gaining access to experiment with powerful and expensive models before the purpose has been considered.

The pattern also minimizes administration: approvals are handled only when the need actually arises, and require no speculative planning or assessment in advance.

Administrators can define default models based on security classification, so that the most appropriate model is preselected. This ensures that new assistants start by default with the most efficient model that meets the requirements.