Communication agent quick start

更新时间:
复制 MD 格式

Intelligent Contact Center provides a communication agent feature that uses large model technology to automate phone calls. These agents replace human operators, engaging customers in natural and fluent conversations. Whether for lead conversion, member reminders, or customer service, the system handles tasks efficiently while significantly reducing your enterprise's labor costs. This guide explains how to start using communication agents.

Intelligent Contact Center also provides other functional modules:

Procedure

image

Using a communication agent involves five main steps:

  1. Create a knowledge base: Upload business-related resources to prepare for agent configuration. These resources include private knowledge files, private voice recordings, and an ASR hotword library.

  2. Create a communication agent: Create a communication agent and wait for it to build.

  3. Configure the agent and media resources: Configure the task's background, objectives, requirements, and high-frequency knowledge. You can also associate a private knowledge base, configure the call voice and volume, handle exceptions, and associate a private voice and hotword library.

  4. Review the communication agent: Submit the agent to Alibaba Cloud for content compliance review. After approval, Alibaba Cloud automatically builds the knowledge. You can then use an authorized number to run a test call to evaluate the overall business effectiveness and the agent's ability to answer key questions. If the performance is unsatisfactory, you can withdraw the agent to edit and optimize its configuration.

  5. Configure a call job: After you confirm the test call results, you can use your approved numbers to start a call job and obtain business outcomes. Create a call job in the console, configure its settings, and import an offline list of numbers for batch calling.

Before you begin

  1. You have an Alibaba Cloud account and completed enterprise real-name registration.

  2. You have enabled Voice Messaging Service.

  3. You have submitted your enterprise qualifications for Voice Messaging Service, and they have been approved. For details, see Submit enterprise qualifications.

  4. You have submitted a script for Voice Messaging Service, and it has been approved. For details, see Add a script.

  5. You have created a service instance. The service instance must be bound to at least one real phone number. For details, see Create a service instance.

Step 1: Create a knowledge base

  1. Create a knowledge base

    On the Intelligent Contact Center console, go to Large Model Communication > Communication Agent > Knowledge Base Management tab. Click Create Knowledge Base, configure the basic information in the dialog box and click OK.

  2. Upload knowledge files

    Note

    After uploading knowledge base files, you can reference them when creating a communication agent.

    1. After creating the knowledge base, click Details on the Knowledge Base Management page.

    2. On the details page, click Create File.

    3. In the Add File dialog box, select Document-based or Q&A-based as the file type, and then click Upload File to select your file.

      Type

      Description

      Example

      Document-based

      Document-based knowledge files are suitable for providing comprehensive, systematic information.

      These are complete documents that contain systematic knowledge, such as product manuals, technical documentation, or operation guides. The system processes these files to extract structured information that users can query. Only .txt files are supported.

      This is test content for a knowledge base. Text uploaded here can be used as knowledge for AI output.
      This is test content for a knowledge base. Text uploaded here can be used as knowledge for AI output.
      This is test content for a knowledge base. Text uploaded here can be used as knowledge for AI output.
      This is test content for a knowledge base. Text uploaded here can be used as knowledge for AI output.
      This is test content for a knowledge base. Text uploaded here can be used as knowledge for AI output.
      This is test content for a knowledge base. Text uploaded here can be used as knowledge for AI output.
      This is test content for a knowledge base. Text uploaded here can be used as knowledge for AI output.
      This is test content for a knowledge base. Text uploaded here can be used as knowledge for AI output.

      Q&A-based

      Q&A-based knowledge files are suitable for providing direct answers to specific questions.

      These files consist of question-and-answer pairs. This format is typically more concise and targeted, allowing users to get answers quickly without reading an entire document. The system uses these Q&A pairs directly to respond to user queries. Files in .csv, .xlsx, and .xls formats are supported.

      The Excel template for Q&A-based knowledge files includes the following columns: Question (Column A), Script 1 (Column B), Script 2 (Column C), and Script 3 (Column D). This allows you to configure multiple scripted responses for each question.

Step 2: Create a communication agent

On the Intelligent Contact Center console, go to Large Model Communication > Communication Agent > Communication Agent Management tab. Click Create Agent, configure the basic information in the dialog box and click OK.

Note

After you create the communication agent, its status on the Communication Agent Management page changes to Building. The build process typically takes 10 to 30 minutes. After the build is complete, the status changes to Debugging.

Step 3: Configure agent and media

  1. Configure voice and call events

    On the Communication Agent Management page, click Voice and Call Event Configuration. In the dialog box that appears, configure settings under TTS Voice Configuration, Hotword Configuration, and Event Handling Configuration.

    Configuration details

    Category

    Parameter

    Description

    TTS voice configuration

    Voice style

    Select a suitable voice style for the agent.

    Speech rate

    Select a speech rate that suits your business scenario.

    Volume

    Select a volume level that suits your business scenario.

    Sample text

    Enter custom text to preview the voice.

    Hotword configuration

    Hotword library

    Select a hotword library. If you do not have one, you must create a hotword library.

    Event handling configuration

    Silence duration

    If no one speaks for the specified duration, the system triggers a silence event. You can set the duration from 3 to 15 seconds.

    Silence detection model trigger

    When the first silence event occurs, the system sends the event to the large model.

    Hang-up on silence

    End the call after a specified number of consecutive silence events. You can set the number from 1 to 5.

    Intelligent answering machine detection

    When enabled, the system automatically detects voice assistants or mailboxes and ends the call, improving call efficiency.

    Maximum call duration

    Sets the maximum duration for a call. The system automatically ends the call when this duration is reached.

  2. Edit the communication agent

    On the Communication Agent Management page, click Edit. Configure the settings on the Basic Information, Agent Customization, Custom Exception Handling, Call Variable Configuration, Primary Intent Configuration, and Output Tag Configuration tabs.

    Configuration details

    Category

    Parameter

    Description

    Basic information

    Business scenario

    Select the scenario that matches your business.

    Agent role

    Define the role of the communication agent.

    Callee role

    Define the role of the person the agent will be calling.

    Language style

    Select a language style that fits your business.

    Associated knowledge base

    Select a relevant knowledge base, such as one containing car model information.

    Agent customization

    Task type

    • Basic Task: Suitable for simple scenarios. Configure an opening remark and task configuration, and clearly describe the task content and flow.

    • Advanced Task: Suitable for scenarios that require breaking a task into multiple sub-tasks with fine-grained control, such as separate opening remarks or key scripts for each. This improves the reliability and stability of task execution.

    Opening remark

    Define the opening remark for the call.

    Task configuration

    • Basic Task: Provide the task instructions for the communication agent.

    • Advanced Task: Break the task into multiple sub-tasks. You need to configure a task name and standard script for each sub-task.

    Process description

    Describe the execution flow and logic of the entire advanced task in detail. Variables are supported. This setting is available only for advanced tasks.

    Custom exception handling

    Custom exception handling

    Configure exception handling policies to ensure the agent responds as you require when encountering specific sensitive questions.

    Call variable configuration

    Call variable configuration

    Manage dynamic variables used during a call. The system automatically adds new variables that you reference in the call configuration. You can customize variable names and types.

    Primary intent configuration

    Primary intent configuration

    Configure the primary intent of the call. The primary intent is the core metric and a required indicator for determining whether the call achieved its goal. The system provides pre-generated templates that you can modify. Currently, primary intents only support structured single-choice output. You can edit the name, description, and possible values for a primary intent.

    Output tag configuration

    Output tag configuration

    Configure optional output tags to capture key information about the call process and outcome, beyond the primary intent. Clearly define the content and criteria for each tag. Output tags can be structured options or text.

Step 4: Submit for review and go live

  1. Submit for review

    After you finish editing the communication agent, click Submit for Review. The agent's status changes to In Review. The review process typically takes about one business day.

    • If approved, the status changes to Ready to Go Live, and you can proceed with test calls.

    • If rejected, the status changes to Build Failed. Review the rejection reason, edit the configuration, and submit the agent for review again.

  2. Expert Debugging and Testing

    Note

    You can add up to 10 authorized numbers and initiate up to 5 test calls simultaneously.

    1. After the communication agent is approved, click Expert Debugging and Testing to perform a test call.

    2. On the Expert Debugging and Testing page, click Add Number.

    3. In the dialog box that appears, bind a test phone number. You must verify the number with a code. Click Confirm to complete the binding.

    4. On the Expert Debugging and Testing page, select one or more numbers and click Test Call. You can select up to 5 numbers for a Batch Test Call.

    5. In the dialog box that appears, click Test Call to start the test.

  3. Go live with the communication agent

    When the communication agent has been approved and its status is Ready to Go Live, click Go Live on the Communication Agent Management page. The agent's status will change to "Online".

Step 5: Create a call job

  1. Create a call job

    On the Intelligent Contact Center console, go to Large Model Communication > Communication Agent > Call Job Management tab. Click Create Call Job, configure the basic information in the dialog box that appears, and then click Submit. After creation, the call job status is Initializing. After you import callee data, the status changes to Running.

    Configuration details

    Parameter

    Description

    Job name

    Enter a name for the job.

    Communication agent

    Select a communication agent that is online.

    Job concurrency

    Specify the maximum number of concurrent calls for this job.

    Note

    Calculation logic: Calls per second + Active calls.

    Associated service instance

    Select the service instance to associate. For details, see Create a service instance.

    Start method

    • Start immediately.

    • Start at a scheduled time.

    Calling days

    Select the days of the week for calling.

    Calling hours

    Select the start and end times for calling within a day. You can add multiple time intervals.

    Redial on no answer

    When enabled, this feature lets you set the number of redial attempts and the interval based on the reason for the call failure.

  2. Import callee data

    1. After creating the call job, click Import Callee Data on the Call Job Management page.

    2. In the dialog box that appears, click Download Template. Fill out the template, and then click Upload File.

      Note

      Template format: Enter the phone numbers of the callees sequentially in the calledNumber column of the template.

    3. Click Confirm to import the callee data.

      Note
      • If you selected "Start immediately" when creating the call job, the job status changes to Running after you import the data.

      • If you selected "Start at a scheduled time" when creating the call job, the job status is Stopped after you import the data.

  3. View call job details

    1. After a call job starts, you can click Details on the Call Job Management page to view its basic information and execution records.

    2. On the call job details page, you can view details for completed, pending, and failed tasks.

Related topics