AI customer service
This document shows you how to enable, configure, integrate, and continuously optimize AI customer service to help you quickly deploy a 24/7 intelligent Q&A service for your website.
AI customer service is an intelligent Q&A product powered by a Large Language Model (LLM). It learns from the knowledge base you provide, such as website content, product documents, and FAQ lists, to automatically provide your website visitors with accurate and instant answers. It is ideal for scenarios like pre-sales consultations, product feature guidance, and website information services, which effectively improves user experience and service efficiency.
|
|
Quick start: Launch your AI customer service
This section explains how to quickly configure and publish your first AI customer service for your website.
Process overview:
Enable the service: Create an AI customer service in the console.
Build the knowledge base: Import website content or upload files to provide the AI customer service with the information it needs to answer questions.
Configure and preview: Customize the appearance of the AI customer service and test its Q&A performance in the preview window.
Publish: Deploy the AI customer service to your website.
Start a conversation: Refresh your website. You will see the AI customer service icon in the bottom-right corner. Click the icon to start a conversation.
Enable AI customer service
Log on to the Wanxiaozi AI Website Builder console.
In the left-side navigation pane, choose .
When you open the AI customer service management interface, the system prompts you to create your first agent.
Build the knowledge base
After you enable the service, the system automatically parses the public content of your website and uses it as the initial knowledge base.
Website content parsing: The system automatically crawls and parses your website content, which forms the basis for the AI customer service's answers. This process may take a few minutes, depending on the size and volume of your website's content.
Supplement knowledge sources: You can also enrich the knowledge base through file upload or by adding an FAQ. For more information, see Knowledge base management.
Configure and preview
Before publishing, customize the appearance of the AI customer service and verify its response quality.
On the Customer Service Configuration page, you can customize the Color Scheme, Service Name, Service Introduction, and Opening Questions to match your brand style.
For detailed configuration instructions, see Style management.
After configuration is complete, ask your AI customer service questions in the Preview window on the right side of the console to test if its answers are accurate and complete.
Once the results meet your expectations, click Publish.
Start a conversation
Refresh your website, and the AI customer service icon will appear in the bottom-right corner. Click the icon to start a conversation.
Knowledge base management
The knowledge base is the information source for the AI customer service. A high-quality and comprehensive knowledge base is key to ensuring effective performance. The system uses Retrieval-Augmented Generation (RAG) technology. When a user asks a question, RAG retrieves relevant information from your knowledge base and uses a Large Language Model (LLM) to generate an accurate answer.
The AI customer service supports three ways to build and enrich your knowledge base:
Website Import: Automatically crawls and parses the content of your specified websites.
File Upload: Upload local files containing structured or unstructured text.
FAQ: Add precise question-and-answer (Q&A) pairs.
NoteWhen a user's question matches multiple knowledge sources, the system prioritizes the standard answer from the FAQ.
Website import
Quickly convert public information from your existing website, such as company introductions, product lists, and news updates, into knowledge by crawling and parsing its content.
Manual refresh: When your website content is updated, you can click Refresh Data on the knowledge base management page at any time. The system will immediately re-crawl your website and update the knowledge base.
Automatic refresh: You can enable the Automatic Refresh feature. Once enabled, the system will periodically and automatically sync the latest content from your website.
After the import is complete, the page displays a list of imported web pages in a table. The table includes columns for Page name, Status (shows "Import successful"), Page size (KB), Update time, and Actions (you can click Delete to remove an imported page).
File upload
Upload local files, such as product manuals, terms of service, and price lists, to supplement the knowledge base with specialized or internal documents that cannot be crawled from your website.
On the knowledge base management page, select the File Upload tab.
Click Upload File and select the files you want to upload.
Supported formats:
pdf,doc,docx,txt,markdown, andexcel.Operational limits: You cannot delete a file while it is being uploaded or parsed. You can only delete it after file parsing succeeds or fails.
After the upload is complete, the page displays a list of uploaded files in a table. The table includes columns for File name, Status (shows the parsing result), Size (KB), Update time, and Actions (you can click Delete to remove an uploaded file).
FAQ
For frequently asked questions, provide standard, definitive answers to ensure accuracy. You can add FAQ items in two ways:
Add Manually
On the knowledge base management page, go to the FAQ management module.
Click Add Question or Add Manually.
In the dialog box that appears, enter a standard Question and the corresponding Answer.
The Question should simulate how a user would ask it. The Answer should be concise, direct, and accurate.
Generate with AI
Click the Generate with AI button.
The system uses AI to automatically generate a batch of recommended Q&A pairs based on your existing knowledge base content (web pages and documents).
You can select the items you need from the generated list and add them to your FAQ list with one click.
Once added, the FAQ list displays all Q&A pairs in a table with columns for Question, Answer, Date added, and Actions. In the Actions column, you can click Edit to modify an existing Q&A pair, or click Delete to remove an unnecessary one.
Service configuration
Style management
You can customize the appearance and conversational behavior of the AI customer service to match your brand style and business needs.
Parameter | Description | Suggestion/example |
Color Scheme | Sets the theme color for the service icon and chat window. | Choose a color that matches or complements your website's brand color. |
Service Name | The name displayed at the top of the chat window. | Example: "Smart Assistant" or "Product Advisor". |
Service Introduction | A brief description of the service's capabilities, displayed below its name. | Example: "I can answer questions about product features and pricing." |
Opening Questions | Guiding questions that the service proactively displays to the user at the start of a conversation. | Purpose: Guide users to start a conversation quickly and showcase the service's core capabilities. |
Contact Information | Alternative support channels for users when the AI cannot resolve their issue. | Can be configured as a phone number, email address, or other contact method. |
Release management
In the preview and debugging interface on the right, you can take the service offline or publish it at any time. |
Analysis and optimization
Continuously monitor and improve the performance of your AI customer service by analyzing service data and user conversations, creating an optimization feedback loop.
View service data
The data dashboard displays key service metrics for the AI customer service within a selected time frame.
In the Wanxiaozi AI Website Builder console, choose from the left-side navigation pane. You can use the time filter in the upper-right corner of the page to select a start and end time to view data for a specific period.
The page contains Service Data and Question optimization tabs. On the Service Data tab, the Question Category area shows the distribution of various question types, including Product Information, Company Information, Pre-sales and After-sales, and Product Recommendation.
Conversation count: The total number of conversations initiated between users and the AI customer service.
User count: The total number of unique users who have initiated a conversation with the AI customer service.
Total questions answered: The total number of questions successfully answered by the AI customer service.
Question resolution rate: (Questions successfully answered / Total questions asked) × 100%.
View conversation history
Review real user conversations to identify gaps in your knowledge base and areas where the AI's responses can be improved.
On the Analysis and optimization > Conversation history page, you can see a list of all user conversations.
Click Details for any record to view the complete conversation, including the user's questions, the AI's answers, and the knowledge base source used for the response.
The list includes fields such as Question Category, Status (for example, Partially answered and Answered), Conversation Time, and Message Count.
Question optimization
When a user asks a question that is not covered by the knowledge base (for example, "How do I buy a domain name?"), the AI customer service replies with a message such as, "The current knowledge base does not contain relevant information, so I cannot provide an answer." The service also guides the user to ask a different question. Below this response, like and dislike buttons allow users to rate the answer quality.
You can use question optimization to quickly add unanswered or poorly answered questions to the knowledge base.
On the Analysis and optimization page, click the Question optimization tab. This list displays all questions marked as "unresolved" by the system or users within the selected date range.
For any unresolved question that is worth adding to the knowledge base, click the adjacent Add Question button.
The system automatically populates the user's query into the Question field of a new FAQ entry. You only need to enter a standard Answer and save the entry.
