Supervised mode behavior

When Zuora AI determines that an action would change data, it presents the proposed action for your review. The action runs only after you approve it. You can decline any proposed action. This per-action approval applies to writes performed through the Zuora AI product experience. Writes performed through Zuora MCP are governed by the approval behavior of the connected third-party AI tool.

Read and analysis behavior is unchanged. Zuora AI can still answer questions and explore data as it does in Read-Only mode.

Supervised mode applies wherever you interact with Zuora AI, including the Side Panel and Full Screen experiences.

Supervised mode remains read and write mode for Zuora MCP. This is because the MCP clients cannot enforce supervised conversations or human-in-the-loop approvals in third-party interfaces the way Zuora AI can in Zuora product experiences.

Zuora records every approved action in the Zuora Audit Trail under the account of the user who approved it. The audit entry is attributed to the approving user, just as if the user made the change directly in Zuora.

Supervised mode does not bypass existing permissions or Controls. You cannot perform any write action that your Zuora role does not allow, and Zuora AI does not perform actions that Zuora AI Controls block.