Create an AI Agent and Train It on Specific Data Sources

Use separate AI Agents for different purposes and "one ring to rule them all"
Written by Konstantine
Updated 2 days ago

Once you've set up your Data Sources (Knowledge Base and Custom Answers), you're ready to create your first AI Agent — a smart assistant that will chat with your customers, find answers instantly, and transfer to a human when needed.

This article will walk you through creating an AI Agent, assigning data sources, and preparing it for action.


🤖 Step 1: Create a new AI Agent

  1. Navigate to Chatbot → AI Agents

  2. Click + Add new AI Agent

  3. Enter a name for your AI Agent
    (Tip: Name it based on its purpose, e.g., “SupportBot”, “Sales Assistant”, or “Legal Helper”)


📚 Step 2: Assign Data Sources

You’ll now choose what data sources the AI Agent will use to generate answers.

  1. Select one or more data sources from:

  2. Drag and drop data sources to set priority:

    • The AI will check sources in order, e.g., check the KB first, then Custom Answers

    • If no result is found, the AI will notify the customer and offer to connect them with a human

  3. Click Proceed to Edit

💡 Best Practice: Use specific data sources for each AI Agent.
For example:
SupportBot → KB: "Troubleshooting" + Custom Answers: "Technical Q&A"
SalesBot → KB: "Features & Pricing" + Custom Answers: "Objection handling"
LegalBot → KB: "Terms & Policies"

⚙️ Step 3: Configure AI Agent's advanced settings (optional)

The Advanced section allows you to shape how your AI Agent sounds, acts, and responds across different languages.

⚠️ Important:
We recommend adding custom instructions only when necessary.
Please write them in English, as it has shown the best results so far. Use prompt engineering best practices.

✨ Personalization

Define the AI's personality.
Example:
You are an AI assistant. Be friendly, clear, and professional.
Or:
Respond to inquiries as if you were Master Yoda from the Star Wars films.

This helps the AI adopt the right tone when chatting with customers.


🧭 Instructions

Guide the AI's behavior with more detailed instructions.
Example:
Help customers with support questions about our app. If the answer isn't found in the data sources, ask if they'd like to talk to a human agent.

This field is useful for defining the agent’s role, fallback behavior, escalation rules, and more.


🌐 Language

Set your AI Agent’s language preference.
Examples:

  • Always respond in the customer’s language

  • Respond only in English

  • If user writes in Spanish, reply in Spanish, otherwise use English

These instructions help prevent unwanted language switching and ensure consistency across multilingual chats.


🚀 Step 4: Add your AI Agent to a Bot Flow

  1. Go to Chatbot → Bot Flows

  2. Create a new flow or edit an existing one

  3. Use the AI Conversation block to drop in your new agent

Remember, AI Agent is just a logical block within your chatbot flow. It’s not a standalone bot - it works as part of the overall chatbot flow element.
You'll find more examples of multi-agent bot flows in AI agents cookbook section

✅ Done! Your Agent is Ready

Once live, your AI Agent will instantly answer customer questions using your selected data sources - no guesswork. And when it’s time to hand off to a human, it does so seamlessly.

➡️ Next up: Analyze AI agent's performance 

Did this answer your question?