You can decide which AI providers, approval modes and script settings people in your organization use with the Agent. The Assistant reads three values from HKEY_LOCAL_MACHINE\SOFTWARE\Policies\Assistant, which you can deploy with Intune or Group Policy.
| Value |
What it decides |
| AIConfiguration |
Which AI providers users can use, and the default model for each |
| ChatPolicy |
Which approval modes users can pick in the Agent chat, and the default |
| ScriptSandbox |
Whether the Agent's scripts in Revit, AutoCAD, Navisworks and Tekla always run in the sandbox |
The full description, with examples you can copy, is on the Assistant portal:
- With AIConfiguration set, AI settings are read-only, and the model menu says "Managed by your organisation".
- With ChatPolicy set to a single mode, the approval menu shows that mode and "Managed by your organisation". A chat saved in a mode that is no longer allowed switches to your default, and the user sees "Your organisation changed this chat to an allowed approval mode."
- With ScriptSandbox set, the sandbox setting in AI settings is locked and says "Managed by your organization".
- A value that cannot be read stops AI requests or new actions, and asks the user to contact their IT administrator. It does not fall back to the user's own settings.