> For the complete documentation index, see [llms.txt](https://docs.omnichat.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.omnichat.ai/en/features/omni-ai-add-on-feature.md). # Omni AI Omni AI can use your knowledge base to answer common inquiries, streamline repetitive support tasks, and provide product suggestions tailored to customer preferences. {% embed url="" %} AI isn't just a trend anymore. It's becoming an essential tool that helps level up your customer service. When used well, it can answer questions faster, lighten your team's workload, boost overall efficiency, and even suggest the right products based on what customers are looking for. Together, these improvements can make a real difference to your bottom line. With Omnichat's upgraded AI feature, you can build your own knowledge base and set up auto-replies for frequently asked questions. This frees up your team from handling repetitive tasks, so they can focus on more complex cases that need a human touch. The chatbot can also suggest products based on customer preferences, helping you improve both support and sales. In this article, we'll show you how to set up your chatbot in Omnichat console so it becomes your team's go-to assistant on the front line. This article shows how to set up Omni AI in Omnichat console to assist with frontline customer service. 1. Upload Knowledge Documents in Knowledge Base * [Access Roles for Knowledge Base](#access-roles) * [Overview](#overview) 2. Create a New AI Agent in Agent Studio * [Access Roles for Agent Studio](#access-roles-1) * [Agent List](#agent-list) * [Agent Categories](#agent-category) * [Service Agent](#service-agent) * [Shopping Agent](#shopping-agent) 3. [Build AI Agent Cards in Chatbot 2.0](#build-ai-agent-cards-in-bot-builder-2.0) * Step 1: [Create a New AI Agent Card and Select an AI Agent](#step-1-create-a-new-ai-agent-card-and-select-an-ai-agent) * Step 2: [Preview and Edit the AI Agent Card](#step-2-preview-and-edit-the-ai-agent-card) * Step 3: [Add the AI Agent Card to the Service Flow](#step-3-add-the-ai-agent-card-to-the-service-flow) * [AI Agent Card Trigger Process & Conversation Handling](#ai-agent-card-trigger-process-and-conversation-handling) * [Notes on Using AI Agent Cards](#notes-on-using-ai-agent-cards) ### Knowledge Base AI works like a human in many ways, and its knowledge base functions as its brain. The information you add to the knowledge base is what the AI uses to understand and answer questions. If the knowledge isn't there, the AI won't be able to respond. But when the content is clear, accurate, and complete, the AI can handle more questions and provide more helpful answers. That's why creating a well-structured knowledge document is important for making your AI assistant truly effective. #### Access Roles Administrator, Manager, Customer Service Manager, Marketing #### Overview

1. Add Document: * Supported file formats: .csv, .doc, .docx, .pdf. Other formats are not accepted. * File size limit: 20MB. * No specific formatting requirements for CSV file contents. * Files with the same name (including extension) will replace the existing ones. * If a file is used by an AI Agent, updates will apply automatically. * AI responses will follow the latest file content.

2. Knowledge Document Name: * Documents are sorted by last updated time, with the most recent ones at the top. * File names are shown in their original names including file extensions. 3. Status: * Processed: The AI has finished processing the file. It is now available for use by AI Agents. * Processing: The AI is still processing the file. It cannot be used by AI Agents yet. * Failed to process: The AI failed to process the file, so it cannot be used. * You can hover over the status to view the error message. Please send the error details to the customer support team or your consultant for further assistance. * If the issue is caused by a system error, please wait until the problem is resolved, then use the reprocess option in the Action menu to upload the file again. 4. Action:

* Download: Allows you to download the original Knowledge Document. * Delete: * If the file is currently used by an AI Agent, it cannot be deleted. The system will show a message indicating that the file is in use. * The message window includes a shortcut link to open the Agent settings in a new tab, making it easier for you to adjust the settings later.

{% hint style="success" %} **Recommended Way to Write a Knowledge Document** 💡 **Service Agent** * Creating a Q\&A set in CSV file: * Column 1: Frequently asked questions. * Column 2: Suggested answers. * You may use extra columns to include extra details, such as question category or reference links. * See the example below for content. 💡 **Shopping Agent** * Creating a product list in CSV file: * You can include product details such as product name, product image (must be a URL), product description, price, and URL. * You may also add extra columns for any other information as needed. * See the example below for content. 💡 **Things to note when uploading a knowledge document:** * The system does not support image recognition and only reads plain text in the file. * To ensure accurate responses, make sure the file includes sufficient and well-structured text content. The more complete the information, the better the AI can respond. {% endhint %} Examples using a CSV file:

### Agent Studio Agent Studio is where you create your AI Agent. Here you can set the tone and wording style, and use clear instructions to guide how the agents interact with customers. They act like a virtual version of yourself. AI Agents will not automatically know what you want it to do. You need to give them clear and specific instructions on what to say and how to say it. The clearer your instructions are, the better the agents will perform according to your expectations. #### Access Roles Administrator, Manager, Customer Service Manager, Marketing #### Agent List

All created Agents are listed here. There are two categories: "Service Agent" and "Shopping Agent".\ The category of each Agent is shown in the top left corner, and you can also tell the difference by the icon. #### Agent Categories

#### Service Agent

1. **Agent Name** (required): The name will be shown in Agent Studio to help you distinguish between different Agents. It will also appear on the AI Agent card.

2. **Knowledge Document** (required): * You can select any previously uploaded documents from the knowledge base using the dropdown menu. There is no limit to the number of files you can select. * If a document is grayed out and cannot be selected, it means the file is still being processed or the processing has failed. 3. **Response Scope** (required): Set clear guidelines for what the Agent is allowed to respond to, including its scope, direction, and limitations. A well-defined scope helps improve the accuracy and relevance of the Agent's replies. You may include the following: * Role definition: Describe the main focus of the Agent’s responses. For example, assisting customers with account setup, answering technical questions, or recommending suitable products. * Service scope: List the types of topics the Agent can handle, such as order tracking, setup guides, product introductions, or user instructions. * Limitations: Clearly specify what the Agent should not answer. For example, it should not provide account numbers, personal information, medical diagnoses, or take actions on behalf of the user. * Avoid vague instructions: Do not simply write "Please refer to the knowledge base." Instead, include a clear topic and enough background context to help the Agent give focused responses. * Examples: * Help answer common questions about purchasing T-shirts from Omnichat Store, such as product details, order status, or return and exchange SOPs. * If the Agent handles questions about health supplements, it can suggest products based on the customer's health condition and lifestyle, such as items for digestion, sleep, or nutrition. However, it should not offer medical diagnoses or treatment advice. For serious symptoms or questions about medication, the Agent should refer the customers to doctors. 4. **Agent's Goal** (required): Set the main objectives of the Agent, such as assisting customers in choosing suitable solutions, answering technical questions, or providing real-time customer support. Clear objectives help the Agent deliver responses that meet expectations. You may include the following: * Tone and wording style: Specify the tone you want the Agent to use, such as polite, neutral, friendly, lively, or professional and concise. * Interaction: State whether the Agent should proactively ask questions and guide customers to provide information. * Specific responses: If needed, set a fixed sentence for the Agent to include after each reply. For example, "I hope the above information has been helpful to you." * Examples: * Begin a conversation by asking the customer about their current condition or issues they want to improve. Then, provide recommendations based on their response, briefly explaining the product's function or suitability. * Use natural and smooth wording to avoid sounding robotic. Avoid repetitive or overly formal phrasing. * Add a reminder at the end of each reply, such as: "If you would like to end the AI assistant flow, please type 'Talk to a human agent' in this chat and we will connect you to a support agent." 5. **Welcome Message** (optional): Welcome message is the first message the AI Agent sends when the AI Agent card is triggered and a conversation begins. 6. **Cancel**: Click "Cancel" and the system will show a reminder that your current edits will be lost if you leave this page.

7. **Save** #### Shopping Agent

1. **Agent Name** (required): The name will be shown in Agent Studio to help you distinguish between different Agents. It will also appear on the AI Agent card.

5. **Save** {% hint style="info" %} Shopping Agent currently does not support content personalization settings. {% endhint %} ### Build AI Agent Cards in Chatbot 2.0 **Applicable Channels:** WhatsApp, Facebook, Instagram, LINE, Webchat (coming soon) #### Step 1: Create a New AI Agent Card and Select an AI Agent Create a New AI Agent Card

Select the AI Agent you created in Agent Studio from the dropdown menu.

#### Step 2: Preview and Edit the AI Agent Card

The AI Agent card consists of three parts: 1. **Welcome Message**: This is the welcome message you set for the card in Agent Studio. If you need to make changes, go back to the card in [Agent Studio](https://console.omnichat.ai/agent-studio/list) to adjust the content. 2. **If the question is within the responding scope**: * The card shows the Agent name you set in Agent Studio. * The right section presents the Welcome Message, Response Scope, and Agent's Goal you set up for that Agent in Agent Studio. * Click "View" in the upper right corner of the Settings section to open the Agent's settings page. * When a customer's question matches the response scope, the system will reply based on the content of the selected knowledge document. 3. **If the question is beyond the responding scope**: * The bot can only connect to a live chat card. * If the customer's question is beyond the response scope or they request a human agent, the bot will move to the "Live chat" card, and the conversation will switch to "Open - Live chat". #### Step 3: Add the AI Agent Card to the Service Flow You can place the AI Agent card at points in the flow that suit your needs. For example, you might add it to a button action in a text card under the welcome message or in the rich menu, so customers can tap it to start an AI Q\&A. You can also set it to trigger with a keyword auto-reply. This way, you can create flexible interactions that suit your needs. #### AI Agent Card Trigger Process & Conversation Handling

1. How to Trigger the AI Agent? * New / Closed / Pending-Bot \ → *(Triggered by customer-defined conditions to activate the AI Agent Card)* \ → AI Customer Service * If a conversation is already bound to an OMO sales, the conversation must first be closed to exit the "Progress" status. Only then can the AI Agent be successfully triggered 2. Rules After Entering AI Chat * If the Chatbot Card is activated via a "Click" action, it will trigger normally. *(e.g., clicking a button on a chatbot card, clicking a LINE Rich Menu, etc.)* * Keyword auto-assign rules will still work. If the conditions are met: * The conversation will move to "My Cases - Progress". * The AI Agent will stop sending automatic replies. * Messages sent by the customer will NOT trigger "Auto-Reply" messages at this time. *(This feature is expected to be supported in 2026 Q3)* 3. How to Exit AI Chat * AI Agent → Closed * Manual Close: Closed manually by a team member. * API Trigger: Closed via a 3rd-party AI Agent calling the Open API - Close case. * AI Agent → Progress (Human Agent) * Manual Follow-up: Followed up by a human agent. * Add WhatsApp contacts and create the conversation. * Auto-Assign Triggered: * Keywords matching the auto-assign rule. * Triggered via a 3rd-party AI Agent calling the Open API - Auto Assign. * AI Agent → Live Chat * Omni AI Escalation: When the AI determines it cannot answer a question (out of scope or unable to understand the query), it automatically prompts a text card with live agent button. Once the customer clicks the button, the conversation is transferred to live chat. * API Trigger: Transferred via a 3rd-party AI Agent calling the Open API - To human API. #### Notes on Using AI Agent Cards The following features do not support using or responding to content with an AI Agent card. To ensure the card works properly, avoid adding it to these modules.

Feature Type	Setup Path
Chatbot	Conditional Reply Card User Input Card Call an API Card
Social Subscription Management	OmniLink Successful Integration Source Channel Message Target Channel Message Integration Failure Message Website Binding Message to customers for successful opt-in Mobile Phone Binding (LINE only) Complete binding message Already bound message Failed to bind message
Advanced Automation	Customer Journey → Message node
Game Modules	Auto messages to winners who received the prize
Saved Reply	AI Agent card cannot be saved as saved reply. Saved reply that contains an AI Agent card cannot be sent.
Auto-close Conversation Reminder Message	Conversation Setting → Action → Edit Content → Reminder message
OMO	OMO Setting → Store Management Auto Close Conversation → Custom by store Binding Message Setting → Binding Success Message → Custom by store Staff Binding Settings → Binding Success Message → Additional Message
Facebook Persistent Menu	Channels → Frequently Used Messages → Persistent Menu
Facebook / Instagram Comment Reply	Reply Settings → Private Message Reply