Skip to content

AI Agent quick tutorial

1. Open the Start Dialog

  1. In your instance, go to Dialogs in the left menu.
  2. Open the ⭐ Start dialog.

Info

⭐ Start dialog is always triggered at the beginning of each conversation.

2. Add a Welcome Message

  1. From the right-side Modules panel, drag a Message module into the workspace.
  2. Add your welcome text. Example:
  3. Connect this Message module to the Dialog Entry input (so the bot knows what to do when a chat starts).

image-20251001-152526.png

3. Add an Agent Module

Now let’s add the AI Agent that will handle conversations.

  1. Drag the Agent module into the workspace.
  2. Insert your prompt. Example prompt structure:

  3. Add Output Paths (topics the Agent should detect and trigger specific actions):

  4. Appointment scheduled → Appointment was successfully scheduled and confirmed.

  5. Escalation/Transfer → Customer requests human support, complains, or asks something not covered.

  6. Call/Chat End → Customer indicates the conversation is finished.

    image-20251001-152805.png

4. Connect the Modules

  1. Connect the Welcome Message to the Agent module (as shown on picture above).
  2. ⚠️ Important: after connecting, click the arrow and press the Pause button.
    → This ensures the bot waits for user input after showing the welcome message.

5. Add output paths logic

Finish the dialog to match your use-case:

  • Process the results if needed (e.g. send result to email, API request)
  • Add Discussion/Call end module to Ending output (Bot will terminate the call or chat)
  • Handle errors and exceptions (Agent module Other output) with message to user and escalation (e.g. send email, create ticket, redirect to human agent)

Your Agent is almost ready! The dialog is already functional now just connect the outputs to complete your logic.

  • Process results – Send collected info (like booking details) via email or API.
  • End conversation – Link the Call End output to a Discussion/Call End module so the bot closes the chat or call properly.
  • Handle errors – Use the Other output (triggers on exceptions or errors) to reply with message or escalate to a human agent, create ticket or send email. Just to be sure, this shouldn’t happen normally.

✅ That’s it! Your AI Agent is now fully operational! Navigate to test page to try it out (message bubble icon at top right corner)

image-20251012-162712.png