Note Agent

更新时间:
复制 MD 格式

Note Agent uses speech recognition and natural language processing technologies to analyze call content in real time. It quickly extracts key information and generates clear, concise summary tags after each call. This makes customer inquiries, feedback, or potential needs immediately available after each call, eliminating the need for manual recording and organization.

Configuration process

image

Procedure

Step 1: Create a Note Agent

  1. On the Call Intelligence > Note Agent Management tab, click Create Note Agent.

  2. In the dialog box that appears, configure the Basic Information and Note Template sections.

    Click Import Template to open the import dialog. You can filter templates by Template ID, Template Name, Industry, and Scenario. Then, in the Actions column, click Import for the desired template to populate its configuration.

    Parameters:

    Parameter

    Parameter

    Description

    Example

    Basic Information

    Note Agent Name

    A custom name for the agent.

    Test Name

    Remarks

    Custom remarks for the agent.

    This is a new Note Agent.

    Note Template

    Note Type

    • call summary: The agent extracts key points from the call, outputs key call information, and generates summary tags after the call ends.

    • call quality inspection: The agent evaluates conversation quality based on your quality inspection requirements and provides a score on a 100-point scale along with suggestions for improvement.

    call summary

    Industry

    • General

    • Education

    General

    Scenario

    A custom scenario.

    -

    Task Background

    Enter the background information for your task.

    For example, in the Task Background text box, you can enter:
    Caller/callee roles: The caller is a customer service representative inviting customers for a test drive, and the callee is a customer interested in buying a car.
    Core task: Invite a customer for a test drive and confirm four key pieces of information: the customer's preferred car model, the specific test drive time, the dealership location, and whether they agree to be added on WeChat.
    Conversation rules: The inquiry flow should be: preferred model → test drive time → dealership location → add on WeChat. Guide the customer to add the dealership's sales representative on WeChat unless they explicitly refuse. For inquiries about discounts, provide a standard response that guides them to add the sales rep on WeChat for details.
    The text box is limited to 2,000 characters.



    Key Information Output / Quality Inspection Requirements

    Key Information Output: If your Note Type is call summary, enter key information tags. Click Add , and in the dialog box that appears, enter a Tag Name and Tag Definition, and select a Tag Type.

    Quality Inspection Requirements: If your Note Type is call quality inspection, enter the quality inspection scoring framework.

    Key Information Output example:

    Configure two output tags: Customer's Preferred Model (Description: "Which car model is the customer interested in?", Type: Text) and Agreed to Test Drive (Description: "Did the customer agree to a test drive?", Values: "Yes" or "No").

    Quality Inspection Requirements example:

    In the Quality Inspection Requirements text box, enter a scoring template for customer service conversation quality. Include five evaluation dimensions: 1. Prohibited sensitive words (20 points), 2. Communication attitude (20 points), 3. Compliance (20 points), 4. Abusive language (20 points), and 5. Complaints (20 points). You must also specify the output requirements: first, provide a total score out of 100; second, explain the reasons for any point deductions; and finally, provide optimization suggestions for dimensions that lost points. A tip below the text box states, "You can modify the scoring framework provided in the dialog box as needed."

    Scoring requirement for quality inspection: If you have multiple inspection items, you must assign a separate score to each. The sum of the scores for all items must be 100.

  3. After you configure the Basic Information and Note Template, click Confirm to create the agent.

    Note

    After creation, the agent's status is building. It changes to ready to go live in about 10 minutes.

Step 2: Test the Note Agent

Note

You can test a Note Agent a maximum of 100 times per day. Once the limit is reached, you cannot perform more tests on that day.

Prerequisites: The status of the Note Agent has changed from Building to Ready to go live.

  1. On the Call Intelligence > Note Agent Management tab, select the Note Agent you created and click Test.

  2. In the dialog box that appears, enter a conversation in JSON format to test the note generation.

    The maximum input length is 5,000 characters. Click Test to generate the result.

    Example:

    [
      {
        "role": "assistant",
        "content": "Hello, is now a good time to talk?"
      },
      {
        "role": "user",
        "content": "Yes, go ahead."
      }
    ]
  3. Click Test to view the result.

Step 3: Go live with the Note Agent

  1. After testing the Note Agent, go to the Call Intelligence > Note Agent Management tab. Select the agent and click Go Live.

  2. In the confirmation dialog box that appears, click Confirm.

Step 4: Subscribe to receipt messages

For more information, see Introduction to receipt messages and configuration process.

Other operations

View details

  1. On the Call Intelligence > Note Agent Management tab, select an agent with the status live or ready to go live, and click Details.

  2. In the dialog box that appears, view the configuration details of the Note Agent.

Take agent offline

Note

Taking a Note Agent offline may affect your business. Proceed with caution.

  1. On the Call Intelligence > Note Agent Management tab, select an agent with the status live, and click Take Offline.

  2. In the confirmation dialog box that appears, click OK.

View version history

  1. On the Call Intelligence > Note Agent Management tab, select an agent with the status live or ready to go live, and click Version History.

  2. In the panel that appears on the right, view the version history.