Google Duplex AI to replace Human Interaction



During Google I/O event 2018, Google announced Glimpse of technology known as Google Duplex.
Duplex is a new Google Assistant’s feature that can do specific tasks for you over the phone. Duplex will be able to reserve a table in your favorite restaurant, also can schedule an appointment, or call any organization to check for their opening hours.
Launch Google Assistant on your Mobile and say, “Book a table for two at Domino’s Pizza at 8 pm today,” and Duplex will do the rest. The AI feature will call the dominos pizza center and make a reservation for you. After the call, you’ll get a notification on your phone confirming the reservation.
This is the power of AI which is making our life a lot easier.

Can we build it with Dialogflow?

Yes, Dialogflow released the feature called “Dialogflow Phone Gateway ”. In its Beta releases, it will provide telephony interface for your mobile.
Currently, you can use only selected phone numbers hosted by google but in near future all public number can be hosted.
Its working in US-English for now.

Let’s get started

  1. Go to Dialogflow Console,
  2. Create New Agent
  3. Delete whatever default Intent you get
  4. Click the settings next to the agent name
  5. Enable Beta features to work with telephony features
  6. Hit the save button

Now let’s set out new intents,

Click on Settings and from the Export/import tab import this agent,


Once you are done with import save your agent again. In the Sidebar you will find integration tab


Click on Integration Tab and there will you find, Dialogflow Phone Gateway,


To configure this click on that and select the language, English is the only available for now. Country is by default US and other countries will be soon there.


By clicking the NEXT you will get choice to select one number to use for whole conversation.


That’s it, click on Create and you are done with the configuration Part.

Let’s test by Calling the number

Enable Small Talk from the Sidebar,
You can now call the number and follow the simple voice prompts. The interactions are defined in your agent.
“You will be asked to say something“ where you can “terminate” the call or “transfer” the call.
Watch a video below to view the conversation I made with an agent.
Though you can test this agent – “+1 650-485-1222

What’s the next step?

This was the simple test Example but what for complex one that showed in google Announcement,
To build something like that here is the Developer Notes,

This feature support reach responses like,

Play audio: It will play given audio in response.

Synthesize speech: You can give response like,

and it will convert it for you to audio.

Transfer call: Transfers the caller to somewhere else where we can say something like handing over a call to some human.

The speech and voice setting can be changed in the setting of your agent in Speech tab.

Voice: You can choose voice generation model and what’s suitable for you, you can go for that.

Speaking Rate: The speaking rate of agent can be changed.

Pitch: Pitch of the voice can be adjusted as per need.

Volume Gain: Audio volume gain can be adjusted.

Audio Effects Profile: Select audio effects profiles that you want to apply to the synthesized voice. Read More here Audio Profiles.

Conclusion : –

The revolution or techs is started with this to stand up odd in the market adopt this kind of features with your business and engage with your customers.
This was a really easy and simple example that can give you simple answers. Let’s see how it goes and will see you all soon in very next blog with complex example and more human-like conversations.


Nikhil Savaliya MEAN Stack Developer

I am Nikhil Savaliya, working as a MEAN Stack Developer at Yudiz Solutions Pvt. Ltd. - a leading mobile app, mobile games and Web Development Company. Technophile, Who wants to explore innovative ideas about AI that can make our life easier.

Comments are closed.