Setting up AI Assist to handle Facebook conversations

The instructions below are for desktops and laptops only.

Use tawk.to’s AI feature to handle your Facebook Business messages automatically, simplifying your workflow with quick, accurate responses and freeing you and your team to work on other tasks.


In this guide, we’ll walk you through the essential steps to set this up and ensure your AI agent is ready to engage with your Facebook audience. 

Connect Facebook and tawk.to

You’ll first need to connect your Facebook Business account to tawk.to. This enables you to receive and handle Facebook messages in your tawk.to dashboard.


Read this helpful guide to learn how to do this: Connecting Facebook Messenger to tawk.to

Enable AI Assist
Next, enable the AI Assist add-on. Learn how to do this with this guide: Getting started with AI Assist
Set up your AI agent 

You can create a single AI agent to respond to chats, tickets, and Facebook messages. You can also set up multiple AI agents to handle specific channels with customized responses.


Learn how to create AI agents with this guide: Deploying multiple AI Agents across support channels


Follow these steps to assign your AI agent to Facebook:

1. Log in to your tawk.to account.


2. On your dashboard, select the property for which you want to set this up. Hover over the property name to ensure you’ve selected the correct one.

3. Click Add-ons on the top menu.

4. Click Settings under AI Assist.

5. Select your AI agent.

6. Go to Channels on your AI agent’s Settings page.

7. Scroll down and enable the toggle for your Facebook inbox.

Your AI agent is now assigned to handle conversations with your Facebook audience.

Customize your Base Prompt for Facebook

You can train your AI agents’ responses by providing clear instructions in the Base Prompt.


If you have multiple AI agents assigned to various channels, you can customize the Base Prompt for each agent and the assigned channel.

Here are some tips to train your AI agent’s responses for Facebook:

  • Keep responses conversational, friendly, and concise.

  • Maintain a casual yet professional tone. Facebook Messenger fosters a more informal communication style.

  • Train your AI agent to offer assistance by introducing themselves as an AI assistant and expressing willingness to help with their inquiry.

  • Mention that your AI agent can answer basic questions, and direct visitors to relevant resources or connect them to a human agent for more complex issues.


Follow these steps to customize your Base Prompt:

1. Go to Base Prompt on your AI agent’s Settings page.

2. Modify the prompt with guidelines for Facebook messages. You can give your AI agent specific instructions to train it to respond appropriately.

3. Click Save when you’ve finished updating the prompt. Your changes are effective immediately.

Here's an example prompt you can adapt to train your AI agent to greet customers when they start a conversation:

- when the customer starts a chat, say
"Hi there! Thanks for reaching out through Facebook Messenger. I'm [AI Agent Name], your friendly AI assistant for [Your Business Name]. Feel free to ask me any questions you might have, or I can connect you with a human agent if needed. Our support team is usually online from 9:00 AM to 5:00 PM, Monday through Friday."

You can customize the Base Prompt to train your AI agent with responses for various scenarios. Learn more about modifying the Base Prompt with this guide: Modifying the Base Prompt


Learn how to guide your AI agent’s responses with this guide: Using Base Prompt to restrict AI Assist's responses. This guide also explains how your AI agent utilizes your data sources to generate responses.


Next, let’s look at data sources and API integrations use to empower your AI agent’s responses.

Data sources and API integrations for smarter responses

To equip your AI agent for success, consider having the following information available:


  • Knowledge Base articles
    Integrate your Facebook Messenger conversations with your existing knowledge base articles. This allows your AI agent to identify relevant articles and suggest them as solutions to customer inquiries.


  • Frequently Asked Questions (FAQs)
    Train your AI agent with your most frequently asked questions specific to Facebook communications. This empowers your AI agent to respond quickly and accurately to common customer concerns.

  • Company policies
    Ensure your AI agent has access to your company policies such as returns, refunds, and shipping policies.

  • Order tracking information
    Empower your AI agent to provide up-to-date order-tracking information.


To learn more about data sources, see this guide: Understanding AI Assist’s Data Sources


Click Data Source to see the data sources available to your AI agent. 

If you have multiple AI agents, you can enable individual data sources for each AI agent.

Click Integration/API to see the available API integrations. You can enable individual API integrations for each AI agent.
Test your AI agent’s responses

Use the built-in test environment to ensure your AI agent’s responses are appropriate for your audience. 

Additional information

Monitor, analyze, optimize

Regularly review your AI agent's performance in handling Facebook conversations. Analyze user feedback to identify areas for improvement. Fine-tune your Base Prompt or FAQs based on user interactions.


Train your team

Inform your live chat agents about the AI agent's capabilities regarding Facebook Messenger messages. This fosters a smooth handoff when the AI agent needs to escalate a conversation to a human representative.


Privacy compliance

Ensure your AI agent handles customer data according to Facebook’s and your business’s privacy policies.


By following these steps, your AI agent will be well-equipped to handle Facebook Messenger inquiries, offering a convenient and efficient customer support experience.


See these guides to learn more about AI Assist:

Using AI Commands

Using Smart Reply

How to fix incorrect responses delivered by AI Assist


If you have feedback about this article, or if you need more help:

Was this article helpful?

0 out of 2 liked this article

Still need help? Message Us