The human‑in‑the‑loop (HITL) settings for AI agents provide users with tools for human oversight of AI agents. Designed to support responsible AI use, these HITL settings allow admins or creators to specify checkpoints for human review, requiring the AI agent to pause and wait for approval before completing a particular action.
By default, AI agent read operations do not require approval. Users can require approval for read actions within the Tool Permissions sections of their settings.
How to configure human-in-the-loop settings in conversations section
Sign in to the Zoom desktop app.
In the navigation menu, click Chat, then click Agents.
Click the desired Agents.
In the right-hand panel under Conversations, select your previous message to the agent, then click the More icon and choose Settings.
Under the Tools permissions for this conversation window, in the section How should this AI agent use tools in this conversation? Select one of the following:
Needs approval for this conversations: Requires user approval before the AI agent performs any action in this conversation.
Always allow for this conversation: Allows the AI agent to perform actions such as editing files, sharing recordings, and sending messages without requiring approval in this conversation.
Custom: Customize whether the AI agent requires approval before performing actions in this conversation, such as editing files, sharing recordings, and sending messages, or can complete them without approval.
Select the app for which you want to edit permissions. In the tools permissions box, click the dropdown menu next to each tool and select one of the following options:
Needs approval: The AI agent always asks for user approval before every action.
Allow for this conversation: Grants permission for the selected AI agent to be used in the current conversation only. Permissions must be granted again in future conversations.
How to configure human-in-the-loop settings in the AI agent section
Sign in to the Zoom desktop app.
In the navigation menu, click Chat.
Click the Agents icon in the top navigation menu.
Under AI agents, click the agent that you want to configure and do one of the following:
In the upper-right corner, click the More icon, then click View settings.
Hover your mouse on the AI agent that you want to configure, click the More icon, and click View Settings.
Hover over the agent avatar you want to configure, and click the gear icon to open settings.
Click the Tool permissions, and select one of the following under How should the AI agent run tools?
Needs approval: The AI agent always asks for user approval before every action.
Always allow: Allow the AI agent to make changes, such as editing files, sharing recordings, and sending messages, without requiring user approval.
Custom: Customize whether the AI agent asks for approval before making changes, such as editing files, sharing recordings, and sending messages, or completes them without requiring approval.
Select the app for which you want to edit permissions. In the tools permissions' box, click the dropdown menu next to each tool and select one of the following options:
Needs approval: The AI agent always asks for user approval before every action.
Always allow: Allow the AI agent to make changes, such as editing files, sharing recordings, and sending messages, without requiring user approval.
Blocked: Prevents the AI agent from performing the selected action.
How to approve AI Agent actions for creating docs and sending emails
Creating document
Sign in to the Zoom desktop app.
In the navigation menu, click Chat and select the Agents icon.
Select an AI agent or existing conversation and send a message prompting the AI agent to create a document.
In the confirmation prompt, choose one of the available permissions.
Allow: Permits the AI agent to create the document.
Deny: Prevents the AI agent from creating the document.
Sending email
Sign in to the Zoom desktop app.
In the navigation menu, click Chat and select the Agents icon.
Select an AI agent or existing conversation and send a message prompting the AI agent to send an email.
In the confirmation prompt, choose one of the available permissions.
Allow: Permits the AI agent to send the email.
Deny: Prevents the AI agent from sending the email.