Help Center | tawk.to | Integrating AI Assist with Twilio calls

Integrating AI Assist with Twilio calls

The instructions below are for desktops and laptops only.

Sometimes, your customers may prefer talking to an agent instead of chatting with a bot. With tawk.to’s AI Assist, you can delight your customers with AI-assisted call escalation.

Using a VoIP service, you can get AI Assist to call your support team when your customers ask to speak to an agent.

In this guide, we show you how to do this using Twilio and AI Assist’s API Integration. You may repurpose the instructions below for other VoIP services, such as Ring Central, Bandwidth or Amazon Connect.

First, set up Twilio

1. Sign up for a Twilio account, or log in if you already have one.

a. Get your Account SID and Auth Token from the Twilio Console.

2. Create a TwiML App:

a. On the Twilio Console, go to Develop > Voice > Manage > TwiML apps.

b. Click Create new TwiML App.

c. Set the Voice Request URL to point to your server endpoint that will handle the call.

d. Take note of your TwiML App SID for later.

3. Create an API Key that authorizes you to interact with Twilio APIs:

a. Go to API keys and tokens.

b. Click Create API key.

c. Take note of the key's SID and Secret.

4. Create a Twilio service:

a. Go to Develop > Functions And Assets > Services.

b. Click Create service.

c. Set the following environment variables:

TWIML_APPLICATION_SID: Your TwiML App SID.
API_KEY_SID: The API Key SID.
API_KEY_SECRET: The API Key Secret.
CALLER_ID: Your Twilio number (include a + and the country code).

5. Configure the dependencies for your Twilio Service:

a. Import the necessary NPM modules for your serverless functions:

6. Create a Twilio Function to issue voice tokens:

a. Go to Develop > Functions and Assets > Functions.

b. Create a /voice-token function to issue voice tokens for authentication.

7. Set up your server:

a. Create a server using a framework like Flask (Python) or Express (Node.js) to handle the outbound call logic.Here’s an example in Python (Flask):

8. Test your setup:

a. Use tools like Ngrok to expose your local server to the internet for testing.

b. Initiate a call by sending a POST request to your /make_call endpoint with the to parameter set to the recipient's phone number.

Create a swagger.json OpenAPI schema

Here’s an example:
Note: Replace “+12345678” with your phone number(s).

Create the script for the API call

Here’s an example:
Note: Replace “+12345678” with your phone number.

Disable AI Assist’s escalation settings

1. Log in to tawk.to, and select the correct property.

2. Click Automation on the left navigation bar.

3. Click Agents in the left submenu.

4. Select your AI agent.

5. Click Settings in the left menu. Under Escalation, disable the When a visitor asks to speak to a human toggle button.

If you don’t see the toggle button, select a shortcut first, and then disable the toggle button.

Complete the integration

1. Click Integrations in the left submenu.

2. Click OpenAPI Server under Custom Tool.

3. This opens the Overview page. Click Install in the upper right corner.

4. Enter the following details:

- The URL of your swagger.json OpenAPI schema in Schema File URL
- The URL of Twilio’s API in API base URL

Then, select the authentication method and click Save.

With these steps, you’ll be able to receive phone calls from AI Assist via Twilio. Remember to test the integration before you deploy. Identify and resolve any potential issues that may affect your customers’ experience.

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

Click the green live chat icon
Schedule a call with us
Visit our community

Was this article helpful?

50 out of 71 liked this article

Still need help? Message Us