Create a Voice AI Agent for Twilio

Hey Rachel, how can we help you today?

Can you help me track my package?

Sure, let me look that up.

Here it is. It looks like your parcel will be delivered in the next two days. The delivery has been delayed because of a storm in the area. Sorry for the inconvenience.

What you just heard was a Voice AI agent built in Kommunicate using the Twilio integration.

Welcome to Kommunicate. In today’s session, we’ll guide you through the process of creating a Voice AI agent with Twilio.

To get started, you’ll need a Kommunicate account. If you don’t have one, it’s easy to sign up. Just go to dashboard.kommunicate.io and click Sign Up.

Since we already have an account, we’ll log in and guide you through the process.

Voice AI agents are particularly useful for deflecting frequent, easy-to-solve customer queries, especially when you don’t have a large customer service team and customers are kept on hold for extended periods.

To begin, navigate to the Bot Integrations section. Here, you’ll see a list of bots that have already been created.

For this tutorial, we’ll create a new bot.

You have the option to create a bot with ChatGPT, Gemini, or Claude. For this tutorial, we’ll use Kommunicate’s own AI platform, known as Kompose.

We’ll name our bot Zeta and choose an appropriate icon.

Depending on your website’s audience, you can also select a different default language. Kommunicate supports over 100 languages, including Arabic, Danish, French, German, and many more.

Once you’ve made your selections, click Save and Proceed.

You also have the option to enable the human handoff feature, where the bot transfers the conversation to a human agent if it cannot answer a particular question. However, we’ll skip this feature for this tutorial and complete the bot setup.

After setup, you can assign all conversations to this bot by clicking here and then clicking Yes.

This will take you to the bot builder screen.

For the knowledge source, you can add documents by clicking here, or you can integrate your knowledge base from Zendesk or Salesforce.

For this particular bot, we’re going to use a website called urbanpiper.com.

UrbanPiper is an Indian company that provides point-of-sale systems, marketing tools for restaurants, and more.

We’ll select their About Us page so that our bot can learn about the company, their blog so the bot can learn about their features, their Contact Us page, and their customer stories.

After selecting these pages, click Train.

While the bot is training, you’ll see the status, and it will continue to train as we proceed.

Next, let’s set the welcome message so we know what the bot will say when it first picks up a call.

The welcome message will be:

Hello from UrbanPiper. How can we help you?

Once you’ve added this, click Save.

Your bot has now learned something new.

You can also set up custom questions and answers so the bot knows how to respond to basic inquiries.

For instance, if someone says “Hi,” the bot can reply, “Hey, how can I help?”

Or, if someone asks for specific documentation, the bot can provide the relevant link so they can directly navigate to that documentation.

If you go back to Knowledge Source and then go to URLs, you’ll see that the bot has already been updated.

As a final step in setting up the bot, go to Settings.

Here, you can choose different AI models again, but we’ll continue using Kompose and add some custom instructions.

Remember, this is a phone call. We want the bot to use a friendly tone and short responses because people don’t want lengthy answers when they’re on a call.

We’ll instruct the bot to recognize the intent behind the customer’s question based on the general topics it was trained on, such as point-of-sale systems and restaurant marketing.

Essentially, we’re giving the bot a specific frame of reference so it understands the context and the kind of answers it should provide.

Click Save, and you’re done.

Now, let’s navigate to Integrations and select Voice Call on the left side of the screen.

In the Voice Twilio Integration section, you’ll find two webhooks.

The first webhook is for incoming calls, and the second webhook is for call status.

How do we use these webhooks?

For that, we need to visit the Twilio Console.

Once there, click View All Numbers and select the number you want to use for your customer service operation.

For this demonstration, we’ll use this number.

You’ll notice two spaces where you can insert the webhooks.

Click and select Webhook here.

When a call comes in, paste the first URL.

When the call status changes, paste the second URL.

You’ve now created your Voice AI agent.

Click Save Configuration to make sure everything is set up correctly.

This is all you need to make your Voice AI agent live.

This Voice AI agent will now be able to answer customer questions at the L1 and L2 support levels.

‍