Building a Voice AI Agent for Feedback Collection
Download
Modo Foco
Tamanho da Fonte
Overview
This document aims to help you quickly start with AI Voice Agent and implement its application in outbound scenarios, providing end-to-end guidance from preparation to call data analysis. You can learn about product features through Introducing AI Voice Agent. This document includes the following content: preparation, research scenario AI Agent setup, outbound task creation, and call data analysis.
Prerequisites
1. Log in to Tencent Cloud and create a Voice application.
2. Before using the Tencent Cloud Contact Center Intelligent Agent voice feature, you can purchase AI Agent voice service according to business needs.
3. Calling an AI Agent outbound requires number access SIP Trunk.
4. After completing application creation and purchase, go to the management panel page.
Research Scenario AI Agent Setup
The process for this scenario is as follows:
1. Create AI Agent: Click AI Agent on the left side of the management console, then click Create Blank AI Agent. Fill in the AI Agent name (for example: research) in the pop-up. The system will automatically create a blank process canvas for you.
2. Build an AI Agent using the invite users to participate in research scenario. Click the start a call node, use the large model to reply. You can refer to the figure below or autonomously set prompt content. Note: prompt content should include the following: persona (AIAgent identity and language style), task (its main feature, the effect you want to achieve), requirements (behavior constraints for the AI Agent).
Add a new Conversation Node as the opening remark. Select AI generation to let the large model generate personalized scripts. Refer to the figure below, set the user reply category for easy participation. For each category, list possible user replies such as: yes, okay. This helps the large model better understand the intent.
3. Set up a survey question inquiry node. Based on user response types, survey questions can be categorized into three types:
If a survey question has a limited number of fixed answers, for example, whether you are satisfied with this service (satisfied/not satisfied), which option you prefer (A or B), and so on, you can configure the answer options in the user reply classification.
If you wish to trace user responses to each question, you can use the Tag Configuration feature (you can view tags in the Post-Call Analysis feature). For example, if a user selects option A for question 1, you can set the tag: question 1-option A. The specific configuration is as shown below:
If a survey question has a large number of fixed answers (configuring each category individually is inconvenient), for example: your new energy vehicle model. You can use the user reply collection feature to configure the content you want to collect.
Path: Choose Collect > Add > Fixed Options. Under this setting, the system automatically collects and tags user responses to such questions.
You can track user responses to this question in post-call analysis.
If the answer to a survey question is open-ended, for example, height, weight, etc., you can use the user reply collection feature to configure the content you want to collect. Under this setting, the system automatically collects and tags user responses to open-ended questions.
Path: Choose Collect > Add > Open Collection.
You can track user responses to this question in post-call analysis.
If collection is not completed (for example, the user indicates it's inconvenient to answer, the user's response does not match the answer you set, and so on), you can choose AI-powered intelligent follow-up or go to the next node. If you want to verify whether the collected information is accurate, you can also configure whether to cross-check the information with the user after collection is completed, as shown in the figure below:
4. During a conversation, a user may refuse to continue participating in the survey at any time. You can refer to the figure below to set a global node and end the dialogue. After the global node is enabled, when any other node triggers the conditions in the global node, the system automatically jumps to this node.
5. To learn more about node function, see node introduction. Once completed, you can test the dialogue effect of the Intelligent Agent. For detailed operations, refer to test dialogue effect.
6. The full canvas view is as follows:
Creating an Outbound Call Task
1. After creating the customer engagement AI Agent, you can create outbound call tasks according to business needs (create a single AI Agent call or create a bulk automatic outbound call task).
2. Variable Substitution: When uploading the called list, you can set variable columns to provide personalized scripts for different users, such as addressing them by surname and gender (such as Mr. Li, Ms. Wang, and so on). When creating the AI Agent, set the variable name you want to replace. The variable format is:
${variable name}, for example, ${name}, ${gender}. You can create different variables based on your business needs.For specific operations, refer to the figure below:
When editing the called list, you need to enter the variable name and variable value, for example,
name: Li, gender: Mr..
Effect: The broadcast script for this node is: Hello, is this Mr. Li? I'm calling to conduct a follow-up with you.
Call Data Analysis
After creating and executing an AI Agent outbound task, you can view the dialogue data, including call connection status, user responses to each question, and more:
Click View Details to go to the specific task details page.
You can view the answer rate of this outbound task and the connection status of each call, and identify the reasons for unanswered calls.
Click Call Details for a specific call record to view detailed information about that call, including Conversation Analytics (post-call Tag status), call flow, and ASR (call transcript).
In the Post-Call Tags section, you can view the user's reply status for that call. As shown in the figure, the user participated in the survey and answered three questions. You can also use the API to query AI Agent call post-call Tag data.
You can also click Batch Export to view all call information for this outbound task.
Related Guidelines
Select the corresponding path based on your deployment method:
First time using? You can deploy a usable AI Agent in just five minutes create a conversation flow or integrate an external model.
Want to learn about our dialogue flow nodes? Go to Dialogue Nodes.
Want to build a knowledge base for your dialogue flow? Go to Build Your Knowledge Base.
How to test your model? Refer to Test Dialogue Workflow.
Want to configure a voice AI Agent with your own model? Go to Integrate External Model, Edit Dialogue Content, Configure Voice Interaction, and Add Call Control.
Want to see application scenarios? Try Voice Notification, Lead Mining, Feedback Collection, or Customer Support.
How to make outbound and inbound calls with an AI Agent? Refer to AI Agent Outbound or AI Agent Inbound.
Want to view data logs? Refer to Call Records, Call Recordings, Latency.
How to create and view post-call Tags? Refer to Post-Call Tags (Created via Dialogue Flow) or Post-Call Tags (Created via External LLM).
Ajuda e Suporte
Esta página foi útil?
Você também pode entrar em contato com a Equipe de vendas ou Enviar um tíquete em caso de ajuda.
comentários