Using Custom Action tool in AI Studio

The Custom Action tool is a built-in tool available in the Tool Templates library in AI Studio that enables a chat agent to perform configurable, AI-driven actions at a specific point in a skill flow, such as summarizing a conversation after an engagement ends, and store the result in a variable for downstream use like updating a CRM.

This is particularly useful when you want the bot to generate dynamic, context-aware outputs without manually scripting every possible response. Beyond conversation summaries, the Custom Action tool supports a range of practical use cases, for example, extracting consumer responses and saving them in a structured JSON schema for further processing or integration with external systems.

Requirements for using Custom Action tool in AI Studio

Table of Contents

How to use the Custom Action tool

  1. Sign in to the Zoom web portal as an admin.
  2. In the top-right corner, click your profile picture or initials, then click Admin Center.
  3. In the side menu, click Product configuration then AI Studio.
  4. Click Virtual Agents.
  5. Create or edit an existing chat agent.
  6. In the Skills section, click Create skill.
  7. Provide a clear description that defines the trigger condition for invoking it. For example, in the Description field, enter, When consumer says custom action, use this skill.
  8. In the Instruction field, specify how the agent should execute the skill. For example, Extract the consumer’s request from the conversation, identify the keyword “custom”, and trigger the corresponding custom action. Return a confirmation message once the action is completed.
  9. Click Add.
  10. To quickly locate this skill later, use the search command “/” and enter custom.

How to use conversation to extract insights

The Custom Action tool can be configured to analyze and extract meaningful insights from ongoing conversations. By defining structured instructions and JSON output keys, you can direct the AI agent to identify and capture specific information such as customer preferences, intent, or purchase details as the conversation continues. Below is an example of how to configure a skill that extracts insights from a conversation.
  1. Sign in to the Zoom web portal as an admin.
  2. In the top-right corner, click your profile picture or initials, then click Admin Center.
  3. In the side menu, click Product configuration then AI Studio.
  4. Click Virtual Agents.
  5. Create or edit an existing chat agent.
  6. In the Skills section, click Create skill.
  7. Fill in the following fields:
  8. In the Instructions section, enter the step-by-step instructions the agent should follow during the conversation. For example:

    Ask the user if they have an account with us. If the user says "yes", ask for their email.
    Collect the following information from the consumer one by one in a conversational manner:
  9. Click Insert or type forward slash (/) to insert a tool.
  10. Fill in the following fields:
  11. In the Instructions section, enter detailed extraction instructions for the tool. Define the JSON keys and their valid values/defaults. For example:

    Extract the Type of Apparel from the conversation.
    The apparel_details is a JSON object with the following Keys:

    apparel_type: Valid options are Polo, Full Sleeve, T shirt. If it is not one of the options, set the key apparel_type field to "Default."
    apparel_size: Extract the size and save it in key apparel_size. Default value is "Medium."
    apparel_color: Extract the color and save it in key apparel_color. Default is "Blue."
    apparel_price: Extract the price and save it in key apparel_price. Default is "$25."

  12. Click Save.

How to use conversation history to generate summary

The Custom Action tool can also be configured to generate a concise summary from the conversation history between the virtual agent and the user. By defining clear summarization instructions, you can direct the AI agent to review the entire conversation and produce a short, plain-text summary that excludes sensitive or harmful content. Below is an example of how to configure a skill that generates a summary from conversation history.

  1. Sign in to the Zoom web portal as an admin.
  2. In the top-right corner, click your profile picture or initials, then click Admin Center.
  3. In the side menu, click Product configuration, then AI Studio.
  4. Click Virtual Agents.
  5. Create or edit an existing chat agent.
  6. In the Skills section, click Create skill.
  7. Fill in the following fields:
  8. In the Instructions section, enter the step-by-step instructions the agent should follow.
  9. Click Insert or type a forward slash (/) to insert a tool.
  10. Fill in the following fields:
  11. In the Instructions section, enter detailed extraction instructions for the tool. For example:


    You are an assistant that summarizes conversations.
    You are given the conversation history:
    [Conversation History]
    {{conversation_history}}
    [End of Conversation History]

    Your task:
    Generate the summary through the conversation history.
    Remember to exclude sensitive information and harmful/offensive content.

    Output format must be STRING only, exactly as follows: "<summary of conversation>"

    Response Format:
    1. Respond only with the summary string.
    2. Do not add any JSON, keys, labels, or explanations.
    3. Ensure the response is a single plain string that can be parsed directly as text.

    Example (format only): <paste an example>

  12. In the Action output section, define the data this action returns when it runs:
  13. Click Save.

Once configured, the virtual agent can automatically summarize the conversation history and store the result in the defined variable for use in subsequent steps or reports.

note icon
This article applies to Zoom Virtual Agent chat agent. If you are using other Zoom Virtual Agent types, such as voice agent or classic chatbot, refer to their respective documentation for setup and channel deployment instructions.