AI Chatbot (Smart Customer Support) Feature Guide

1. Overview

AI Chatbot is Webhub's AI-powered customer support tool integrated into Copilot (your website admin). It improves visitors' consultation experience and boosts service response efficiency.

With AI Chatbot, you can:

• Upload knowledge documents and app data for the AI to learn: Train the bot to answer visitors based on the product docs, FAQs, and business rules you provide.
• Configure the chat widget's appearance and the AI's reply copy: Customize the chat window UI, welcome message, no‑match (fallback) reply, and more—so the experience matches your brand.
• Configure which pages the AI customer support widget appears on: Choose where to show the chat entry (e.g., Home, Product, Contact) to avoid distracting users or missing key pages.

After proper configuration, visitors can start a chat anytime while browsing your site. The AI will respond automatically using the knowledge it has learned—reducing pressure on human agents and improving response speed.

2. Features

2.1 Upload Knowledge

2.1.1 Overview

The quality of AI Chatbot's answers depends on whether your knowledge base is complete and accurate. In this module, you can add knowledge in two ways:

● Upload File Knowledge: Upload documents such as PDF, Word, and TXT. The system will parse the content and add it to the knowledge base.

● Upload App Knowledge: Sync data from existing site apps (such as Products, Articles, Forms, etc.) into the knowledge base so the AI can reference it when answering.

You can use both methods at the same time. The AI chatbot will answer visitors by combining information from your documents and your app data.

2.1.2 Upload File Knowledge

2.1.2.1 Overview

Import written materials—such as product manuals, FAQs, and policy notes—by uploading local files or online documents. This enables the AI to answer visitors based on those materials.

Supported formats: Usually includes PDF, Word (.doc/.docx), TXT, and more. Please refer to the description shown in Copilot under the "Upload File" entry for the exact supported formats.

2.1.2.2 File Parsing

• Steps: In "AI Chatbot-Knowledge-Documents", choose a file and upload it. The system will parse the document (extract text, recognize structure, etc.).
• Parsing process: After uploading, the job enters the parsing queue. Copilot will show statuses such as "Parsing in progress" or "Waiting to parse". Large files or complex layouts may take longer.
• Note: Make sure your file contains extractable text. Scanned PDFs or image-only content may not be parsed correctly. We recommend using documents where text can be copied.

2.1.2.3 After Parsing Completes

• After parsing succeeds, the document content is written into the knowledge base and can be used to answer visitors' questions.
• You can view uploaded files and their statuses in the knowledge management list. If parsing fails, follow the prompt to check the file format or upload again.
• If the document is updated later, you can upload the updated file (with the same name or a new name). Depending on the product behavior, it may either replace the existing knowledge or append new content.

2.1.3 Upload App Knowledge

2.1.3.1 Overview

"App Knowledge" means syncing structured data from the apps/modules already in your website (e.g., Product Center, News/Articles, custom data lists, etc.) into the AI Chatbot knowledge base. This allows the AI to answer questions like "What products do you have?" or "What is this article about?" using up-to-date data that is synced in real time or on a schedule.

2.1.3.2 Select Apps

• In the "Upload App Knowledge" or "App Knowledge" configuration page, from the list, select the apps you want to sync (e.g., "Products", "Articles", etc. The available app types depend on what you see in Copilot).
• You can select multiple apps. After you save, the system will pull fields such as titles, summaries, and details from those apps according to predefined rules and use them as knowledge sources.

2.1.3.3 Wait for Parsing

• After you select apps and save, the system will parse and index the app data. The status may show "Parsing" or "Waiting to parse".
• If the data volume is large, parsing may take several minutes. Please wait, or refresh the page later to check the status.

2.1.3.4 After Parsing Completes

• After parsing completes, the app data is added to the knowledge base. The AI can then answer visitors based on this content.
• If scheduled syncing is supported, the knowledge base will update automatically (or according to the plan) when the app content changes—so you don't need to re-upload every time.

2.2 Configure the Chat Widget

2.2.1 Overview

The chat widget is the interface visitors use to interact with the AI. In this module, you can configure how the chat widget appears, what it says, and how it behaves—so it matches your brand and is easy to use.

Main settings include: Integration Settings (e.g., display scope, on/off), Message Settings (welcome message, no-match reply, etc.), Feedback Settings (e.g., satisfaction collection), Style Settings (position, color, size, etc.).

2.2.2 Integration Settings

Used to control the AI customer support widget on your site, including its display scope and basic behavior, for example:

• Enable/Disable: Whether to show the chat entry on the live site.
• Display pages: Choose which pages display the chat widget (e.g., entire site, only Home, only Product & Contact). This helps avoid interrupting users on unnecessary pages, and ensures the entry is present on key pages.
• Devices: Show on desktop only, mobile only, or both—depending on your operating strategy.
• The actual options may vary based on what is available in Copilot.

2.2.3 Message Settings

Used to set fixed copy and fallback replies in conversations, for example:

• Welcome message: The first message a visitor sees when they open the chat. It can include a greeting, a short description of what the bot can help with, or guidance on what to ask.
• No-match reply: The automatic prompt shown when the AI can’t find a suitable answer in the knowledge base (e.g., "I don't have an answer for that yet. Please rephrase or contact a human agent.").
• Human handoff prompt: If handoff to a human agent is supported, you can configure the message shown before transferring.

Consistent and friendly copy improves the visitor experience and reduces misunderstandings.

2.2.4 Feedback Settings

Used to collect visitor feedback after the conversation ends or after the AI replies, for example:

• Enable feedback: For example, thumbs up / thumbs down, or a rating for "Was this answer helpful?"
• Feedback copy & options: For example, "Helpful / Not helpful" or a 1–5 star rating, which helps you optimize the knowledge base and response strategy.

You can view feedback data in Copilot reports or related modules and use it to continuously improve the AI customer support experience.

2.2.5 Style Settings

Used to adjust the visual presentation of the chat widget so it matches your website style, for example:

• Position: Where the chat entry appears on the page (e.g., bottom right, left side).
• Theme color / background color: Colors for the main button, chat window background, border, and more.
• Size: Chat window width/height, corner radius, etc.
• Icon & label: The icon for the entry button and the helper text shown on hover/expand.

After you save, the live chat widget will update to the new style. We recommend checking it in “Preview" before publishing.

2.3 Preview

2.3.1 Overview

Before you open the AI Chatbot to visitors, use "Preview" in Copilot (or a simulated environment) to check how the widget looks and how it answers—so you can quickly fine-tune your settings.

2.3.2 Ask Test Questions

• In the preview screen, you can type business-related questions and send them to check whether the AI pulls the right information from the knowledge base and answers as expected.
• We recommend testing these types of questions:

○ Questions that are clearly covered in the knowledge base (should receive accurate answers);

○ Ask the same question in different wording (to check generalization);

○ Niche questions that are outside the knowledge base (should trigger the no‑match reply or human handoff logic).

• Based on the preview results, go back to modules like "Upload Knowledge" or "Message Settings" to add or adjust content. When you’re satisfied, use "Integration Setting" to officially make it available to visitors.

2.3.3 View Chat History

3. Frequently Asked Questions

3.1 What file types are supported? Are there size or quantity limits?

Usually, formats with extractable text are supported, such as PDF, Word (.doc/.docx), and TXT. For the exact supported formats and single-file size / total quantity limits, please refer to the notes shown in Copilot on the "Upload File Knowledge" page. We recommend keeping each file reasonably small and using text-based documents rather than scanned PDFs to improve parsing success.

3.2 After uploading or updating knowledge, how long does it take to take effect?

• File knowledge: It becomes available as soon as parsing completes—usually within a few minutes.
• App knowledge: It takes effect after parsing/sync completes. If scheduled syncing is supported, it will update according to the sync frequency.

If you don't see changes after a long time, check that the status shows "Completed", and verify by asking questions in "Preview".

3.3 How can I make the AI’s answers more accurate?

• In "Upload Knowledge", continuously add documents and app data that are highly relevant to your business, and keep the content accurate and clearly written.
• In "Message Settings", write a good "No‑match reply" to guide visitors to rephrase or contact a human agent—so the AI doesn't guess.
• Review feedback data and common missed questions regularly, then add knowledge or refine wording in a targeted way.

3.4 Can I hide the chat widget on certain pages?

Yes. In the "Configure Chat Widget" module under Integration Settings, you'll typically find options like "Display Pages" or "Display Scope". You can choose to show the chat entry only on specific pages (e.g., Home, Product, Contact) and hide it on other pages.

3.5 Will answers in Preview match what visitors see on the live site?

With the same knowledge base and settings, Preview and the live site usually use the same model and knowledge base, so the answering logic should be consistent. If there is admin-only test data or settings that haven't been published, results may differ slightly. We recommend validating the final experience on the live site after publishing.