How to Get ChatGPT to Make Phone Calls For You (Custom GPT Tool)

Liam Ottley
9 Apr 202408:06

Summary

TLDRThis tutorial demonstrates how to set up a custom AI assistant using chat GPT and Bland AI to make phone calls. The process involves creating a GPT account, crafting a detailed prompt for calling objectives, and integrating Bland's API for outbound calls. The AI assistant is programmed to gather necessary information, construct the appropriate task prompt, and execute the call, also providing a transcript of the conversation.

Takeaways

  • πŸ€– The tutorial demonstrates setting up a custom AI service that can make phone calls on a user's behalf.
  • πŸ’‘ The process involves creating a GPT account and using an AI calling service integrated with a custom GPT.
  • πŸ“‹ A clear objective and a phone number are required to successfully make a call.
  • πŸ”— The tutorial references a free school community where users can access the prompt and other resources.
  • πŸ› οΈ The script outlines the technical steps of configuring the GPT with the necessary API from Bland AI service.
  • πŸ“ The importance of crafting a detailed task prompt is emphasized for effective communication during the call.
  • πŸ”„ The script explains how to use the Bland API for both initiating calls and retrieving call transcripts.
  • πŸ”‘ API keys from Bland are required for authentication and are added in the GPT action setup.
  • πŸ“Œ The tutorial provides an example scenario of scheduling a lunch meeting with a business partner.
  • πŸ“ˆ The AI assistant confirms the appointment and retrieves a transcript of the call for future reference.
  • πŸŽ“ The video encourages viewers to explore the prompt and schema provided in the school community for further learning.

Q & A

  • What is the main purpose of the tutorial?

    -The main purpose of the tutorial is to demonstrate how to set up Chat GPT to make phone calls on behalf of the user.

  • What is required to successfully make a call using the AI service?

    -To successfully make a call, two things are required: a phone number to call and a clear objective from the call.

  • How does the AI calling service interact with the user to gather information?

    -The AI calling service takes in a phone number and the user's objective for the call. If it needs more information to complete the task, it will ask the user for that information.

  • What is the role of the custom GPT in this process?

    -The custom GPT's role is to generate a task based on the user's instructions, which includes the phone number and the purpose of the call, and then use this information to make the call through Bland AI's service.

  • What is Bland AI, and how does it relate to the AI calling service?

    -Bland AI is a service that enables the sending of outbound phone calls. It is connected to the custom GPT, which uses Bland AI's capabilities to make phone calls based on the user's requests.

  • How does the user provide the necessary information for the AI to make a call?

    -The user provides the necessary information by giving the AI a clear objective for the call, such as making a restaurant reservation or scheduling an appointment, and providing the relevant contact details.

  • What is the significance of the task variable in the Bland API?

    -The task variable is significant because it contains the full prompt of what the AI assistant needs to do, such as reaching out to book an appointment or ask a specific question over the phone.

  • How does the AI ensure it has the correct information before making a call?

    -The AI ensures it has the correct information by asking the user for any additional details it needs to successfully complete the call, such as the specific time or location for a meeting.

  • What is the process for retrieving a transcript of the phone call?

    -After the call is completed, the AI can use the phone call ID to retrieve a transcript of the conversation through Bland AI's API, providing a summary of what happened during the call.

  • How can users access the prompt and schema used in the tutorial?

    -Users can access the prompt and schema by signing up for the creator's school community, where they will find the prompt and related materials available under the YouTube tab.

  • What is the potential application of this AI calling service?

    -The AI calling service can be used as a personal assistant to make calls for users or integrated into systems for outbound calling capabilities in various applications, such as customer support phone lines for businesses.

Outlines

00:00

πŸ€– Setting Up AI for Phone Calls

This paragraph introduces the concept of using an AI service to make phone calls. The speaker explains how to set up a chat GPT to initiate phone calls by connecting it to an AI calling service. The process involves creating a custom GPT with a specific prompt, which defines the AI's role as a personal assistant responsible for making calls on behalf of the user. The speaker details the need for a phone number and a clear objective for each call, and how the AI will request additional information if necessary. The paragraph outlines the technical steps of integrating the AI with Bland AI's outbound call functionality, emphasizing the importance of the task variable in crafting the call prompt according to Bland's documentation.

05:01

πŸ“ž Testing the AI Call Service

In this paragraph, the speaker demonstrates the practical application of the AI call service by conducting a test call. The AI is instructed to call the speaker's business partner, Josh, to arrange a lunch meeting. The speaker provides specific details such as the preferred day, time, and location for the lunch, as well as alternative options. The AI then generates a detailed task prompt based on the provided information, which is used to guide the call. The paragraph concludes with a successful call demonstration, where the AI confirms the lunch appointment with Josh and provides a call transcript, showcasing the AI's ability to not only make calls but also to provide a summary of the call's outcome.

Mindmap

Keywords

πŸ’‘AI calling service

AI calling service refers to the use of artificial intelligence to automate the process of making phone calls. In the context of the video, it is a technology that allows the AI to dial a number and communicate with a person on behalf of the user. The AI calling service is integrated with a custom GPT (Generative Pre-trained Transformer) model to execute tasks such as scheduling appointments or asking questions over the phone. An example from the script is the AI calling the user's business partner to confirm lunch plans.

πŸ’‘Chat GPT

Chat GPT is an AI model designed for conversational interactions. In the video, it is used as the interface through which the user interacts with the AI calling service. The user inputs their request into Chat GPT, which then generates the necessary information and instructions for the AI to make the phone call. It serves as the user's personal AI assistant, taking in the user's instructions and converting them into a format that the AI calling service can understand and execute.

πŸ’‘Custom GPT

A custom GPT refers to a modified or tailored version of the Generative Pre-trained Transformer model, which is designed for specific tasks or purposes. In the video, the custom GPT is set up to work with an AI calling service to make phone calls on behalf of the user. It is programmed with a specific prompt that outlines the role of the AI as a personal assistant, tasked with making calls to various entities such as restaurants, families, or businesses.

πŸ’‘Phone number

A phone number is a series of digits that uniquely identifies a subscriber's line for communications over a telephone network. In the context of the video, it is essential for the AI calling service to have a valid phone number to initiate an outbound call to the desired contact. The user provides the phone number as part of the information required to complete the call task.

πŸ’‘Clear objective

A clear objective refers to a well-defined goal or purpose for a particular action or task. In the video, it is crucial for the success of the AI calling service to have a clear objective from the user, such as scheduling a lunch meeting or confirming an appointment. This objective guides the AI in generating the appropriate conversation script and ensures the call achieves its intended purpose.

πŸ’‘Bland AI

Bland AI appears to be a fictional service in the video that provides the capability to make outbound phone calls. It is integrated with the custom GPT to enable the AI to send calls on behalf of the user. The service requires specific information, such as the phone number to call and the task to be executed, to successfully make a call.

πŸ’‘API

API, or Application Programming Interface, is a set of protocols and tools that allows different software applications to communicate with each other. In the video, the user configures the custom GPT to interact with Bland AI's API, which facilitates the process of making phone calls. The API enables the GPT to send the necessary information, such as the phone number and call objective, to initiate and manage the call.

πŸ’‘Task prompt

A task prompt is a structured message or script that outlines the specific task the AI is expected to perform. In the video, the task prompt is generated by the custom GPT based on the user's instructions and is used to guide the AI calling service in its interaction with the call recipient. It includes details such as the purpose of the call and the desired outcome.

πŸ’‘Transcript

A transcript is a written version of spoken words, typically used to record and review the content of a conversation or presentation. In the video, after the AI calling service completes the call, a transcript is generated and provided to the user. This transcript serves as a record of the conversation that took place during the call, allowing the user to review the details and outcome of the interaction.

πŸ’‘Custom action

A custom action refers to a specific function or process that is created to meet particular needs or requirements. In the video, the user sets up a custom action within the custom GPT that utilizes Bland AI's capabilities to make outbound phone calls. This action is tailored to the user's needs, such as making calls for scheduling appointments or asking questions.

πŸ’‘School Community

The School Community mentioned in the video is a platform or group where users can share resources, knowledge, and support each other. It is where the user can find the detailed instructions, prompts, and other materials related to setting up the AI calling service and custom GPT. This community provides a space for users to learn and engage with the content of the video more deeply.

Highlights

The tutorial introduces a method to set up Chat GPT to make phone calls.

An AI calling service is connected to a custom GPT for automated communication.

The AI can reschedule appointments, create appointments, ask questions, or confirm details.

A Chat GPT Plus account is required to create a custom GPT.

The custom GPT is named 'my AI phone assistant' for the purpose of the tutorial.

The AI's main role is to make calls on behalf of the user to various entities such as restaurants and businesses.

To make a call, the AI needs a phone number and a clear objective from the call.

The tutorial provides a prompt template that can be used for the AI to understand the call's purpose.

Bland AI is used for sending outbound phone calls and integrating it with the custom GPT.

The process involves filling out variables such as phone number, task, and 'from' on Bland AI's platform.

The tutorial explains how to use Bland's API to make an outbound call and retrieve a transcript.

The API key from Bland is required for authentication and is added in the custom GPT's settings.

The tutorial demonstrates how to instruct the GPT to create a task prompt for each call.

The structure of the task prompt is based on the Bland documentation and the AMHBL method.

The AI can confirm details and provide a summary of the call using the fetch transcript tool.

The tutorial encourages users to explore the prompt and schema for custom GPT creation.

The video concludes by suggesting potential personal and business applications for the custom GPT tool.

Transcripts

play00:00

hey guys today I have a super quick

play00:01

tutorial for you showing how to set up

play00:02

chat GPT to make phone calls for you I'm

play00:05

calling to ask if you'd be interested in

play00:07

having lunch with Liam next Monday and

play00:08

just s off yes we're going to be using

play00:09

an AI calling service and connecting

play00:11

that to a custom GPT so that we could

play00:13

tell chat GPT hey here's this phone

play00:15

number can you please call them and do

play00:17

this it's going to be able to reach out

play00:18

and reschedule an appointment or create

play00:20

an appointment or ask them a question or

play00:22

confirm something so really cool let's

play00:24

jump into it so to start things off

play00:25

you're going to see a chat gbt Plus

play00:27

account it's going to allow you to

play00:28

create your own gbt so if we go to the

play00:30

explore gpts tab here create a new one

play00:32

I'm going to call this one my AI phone

play00:37

assistant call on my behalf for the

play00:40

prompt I'm going to paste in the one

play00:41

that I've created for this video it's

play00:42

quite a complex one but you guys are

play00:43

going to be able to get it on my free

play00:44

school Community down below so you can

play00:46

sign up to school you'll be able to find

play00:47

at the YouTube Tab and find this

play00:49

particular video and there'll be this

play00:50

prompt and everything related to this

play00:51

video will be available there so once

play00:53

we've got the prompt in there it's

play00:54

basically just saying that you are my

play00:55

personal AI assistant your main role is

play00:56

to make calls on my behalf to

play00:58

restaurants families businesses Etc in

play01:00

order to successfully make a call you

play01:01

need two things a phone number and a

play01:03

clear objective from the call depending

play01:05

on the objective of the call make sure

play01:06

you've collected sufficient information

play01:07

from me so that you can actually

play01:09

successfully complete the task so it's

play01:10

supposed to take in a phone number take

play01:12

in what I'm try to do on the phone

play01:14

number say call this person and do this

play01:16

and then if it has any questions it will

play01:17

ask me for any information that it

play01:19

thinks it needs to complete it

play01:20

successfully a conversation started is

play01:22

call someone for me uh we can leave we

play01:24

browsing on I don't think we'll need

play01:26

image generation then we get to the real

play01:28

meat and potatoes of this thing which is

play01:29

to create a custom action that using

play01:31

Bland AI so next step is to go to bland.

play01:34

a and create an account if you haven't

play01:36

already what we're trying to do within

play01:37

our GPT is to take this send call

play01:40

capability that bland AI has to send an

play01:42

outbound phone call and pull it into our

play01:44

GPT so that our chat GPT our custom GPT

play01:48

can fill out the information for us so

play01:49

here you can see on the right it might

play01:51

be a little bit technical for some of

play01:52

you but stay with me in order to

play01:53

successfully make an outbound phone call

play01:55

with BL AI we need to fill out some of

play01:57

this data here that you can see on

play01:58

screen so we need a phone number we need

play02:00

the task and we need the from Etc so

play02:02

these variables are what we're going to

play02:04

be filling with our GPT on the other

play02:06

side every time we want to make a call

play02:07

when you're on Bland itself and you fill

play02:09

out this information here and then click

play02:10

Send call we're going to be taking this

play02:12

exact same functionality by getting the

play02:13

GPT to fill out this form essentially

play02:16

and send a call using Bland's API the

play02:18

process of converting this into the open

play02:20

API schema that is required to provide

play02:22

it as a tool within your GPT is the

play02:24

tricky part but I've done this off

play02:25

camera for you so it'll be available on

play02:27

the school Community as well you want to

play02:28

copy that and just past P it in here and

play02:30

I can take you through it roughly a

play02:32

quick overview of what we're doing here

play02:33

so this is the Bland API and we're

play02:35

providing a URL that our GPT is going to

play02:37

be able to call and then we have a bunch

play02:39

of the variables that we need to pass in

play02:41

the schema has two different

play02:42

functionalities firstly it allows us to

play02:44

initiate a phone call so send off that

play02:46

information okay I want to call this

play02:47

person and I want to do this which we've

play02:49

just gone over on the Bland

play02:50

documentation there but it also has the

play02:52

ability to retrieve a transcript of the

play02:54

phone call so after the call has been

play02:56

completed we're going to be able to take

play02:57

that same phone call ID and say hey what

play02:59

actually happen in that phone call tell

play03:00

me what happens so that I'm on the same

play03:02

page if we scroll down a little bit

play03:03

further you can see the properties that

play03:05

are required to initiate a phone call so

play03:07

obviously we're going to need a phone

play03:08

number that we're going to be calling

play03:09

and then most importantly we have the

play03:11

task which is the purpose or objective

play03:13

of the phone call now this is the most

play03:14

important part for us because we're

play03:16

going to need to get the gpt2 based off

play03:18

our instruction saying I want you to do

play03:20

this it's going to need to generate a

play03:22

task a big prompt to send away to this

play03:24

phone call to to send away to Bland in

play03:26

order to actually successfully execute

play03:27

what we wanted to do before we jump into

play03:29

the nitty-gritty of getting the task

play03:30

probed coming out correctly we need to

play03:32

finish off creating our action here to

play03:34

provide our authentication to use the

play03:35

Bland API we need to head back to Bland

play03:37

go down to our account down the bottom

play03:39

we go API Keys API key we can copy it

play03:45

head back to chat gbt click on the Cog

play03:47

next to authentication go API key paste

play03:50

in your API key and then click on custom

play03:52

and we need to enter

play03:55

authorization and with that we can click

play03:57

save and then we've set up our Bland AI

play03:59

tool we can go back let's give it a

play04:01

little uh darly generated image we

play04:04

almost ready to give this thing a spin

play04:05

now but I want to take you back to the

play04:06

instructions here to show how we're

play04:08

instructing the gbt to correctly create

play04:11

the prompt or the task that we're

play04:12

sending off for every call so as I

play04:14

mentioned the Bland calls the outb

play04:16

Clause that we make with Bland require a

play04:17

task variable that is going to be the

play04:19

full prompt of UR this your job is to do

play04:22

this um you are supposed to be reaching

play04:23

out in order to book an appointment with

play04:25

this Etc So the instructions need to go

play04:27

into the task section and what I've done

play04:28

this is based off the the Bland

play04:30

documentation itself if you go to

play04:31

documentation here and the am hlb

play04:34

prompting guide all of this

play04:35

documentation here is something that I

play04:37

use in my voice agents regularly it is a

play04:39

very good structure and way of layout to

play04:41

get these things working correctly so if

play04:42

you want to take a look at this it's

play04:43

available on the documentation but I've

play04:45

taken everything from this and I've

play04:46

converted it into a prompt that's going

play04:48

to allow our custom GPT to generate

play04:50

these kind of specific prompts every

play04:51

time it sends a phone call as you can

play04:53

see down here I've got creating task

play04:54

prompts using the strict am hlb method

play04:57

and down here it explains how the task

play04:59

prompt should be structured gives an

play05:01

example script you can head down and get

play05:03

there from the school Community if

play05:04

you're looking to follow along but this

play05:05

prompt is really the most important part

play05:06

of this build alongside the open API

play05:09

scheme that's available on school as

play05:10

well right so now we're ready to give

play05:11

this thing a spin so hi can you please

play05:15

call someone for

play05:18

me it's asking for the phone number who

play05:21

I'm going to be calling and then the

play05:22

specific objective or message they like

play05:24

me to convey can you please call my

play05:26

business partner Josh and ask him if he

play05:27

wants to do lunch next Monday dle

play05:30

thought around 12:00 p.m. would be good

play05:33

but I am flexible within an hour each

play05:36

side I can only do

play05:38

mondy thought it's going to ask the

play05:40

phone number also is there a specific

play05:43

place and do off that you'd like to go

play05:44

for lunch so I can say yeah Elmo's

play05:47

Bakery is good and his phone number is

play05:50

right so now I should be able to get a

play05:52

phone

play05:53

call on my handy dandy phone here see

play05:57

it's going to pause while it writes that

play05:58

big long task prompt as we' specified in

play06:01

our instruction so I can pause for a

play06:02

little bit make sure it's written out

play06:03

quite a lot and you'll see just how much

play06:05

it's filled into that task prompt to

play06:07

really guide it and in terms of how it

play06:08

should be behaving on this phone call so

play06:10

we can pop this down here there we go I

play06:12

Josh bis Nino Liam's assistant we have

play06:14

this full prompt that's been written out

play06:16

which is in line with the Bland

play06:17

documentation on how to write these

play06:19

prompts so now I should be able to

play06:20

confirm this and I will get a call here

play06:23

on my

play06:25

phone hi Josh this is Nina Liam's

play06:28

assistant I'm calling to ask if you'd be

play06:30

interested in having lunch with Liam

play06:31

next Monday in dor he suggested Elmo's

play06:34

bakery around 12:00 p.m. but he's

play06:36

flexible up to an hour either side he

play06:38

mentioned he can only do Monday do you

play06:41

think that would work for you yeah could

play06:43

we do 1 p.m. Liam can accommodate within

play06:46

an hour of the proposed time so 1: p.m.

play06:48

at Elmer's Bakery will work perfectly he

play06:50

looks forward to meeting you there

play06:52

perfect thank you bye and there you go

play06:54

we've been able to successfully send a

play06:55

call from our gbt to the phone of in

play06:57

this case my business partner I'm asking

play06:59

me to confirm something so this is

play07:00

obviously just a start um I'm not trying

play07:02

to show you all the different use cases

play07:03

in this video I'm just showing you how

play07:04

you can create a custom tool for your

play07:06

gpts for your custom gpts whether you

play07:08

want to use this as a personal assistant

play07:10

or whether you want to build this into

play07:11

your systems and start to use these

play07:13

outbound calling capabilities within

play07:14

your build as well and as you can see

play07:16

here it's actually used the other fetch

play07:17

transcript tool that we provided when we

play07:19

set up our tool and it's given me a

play07:21

breakdown I was saying I've successfully

play07:23

contacted Josh and he's agreed to have

play07:24

lunch with you and the next Monday in

play07:26

dorf he requested it could be at 1: p.m.

play07:28

at almma Bakery Bas on your flexibility

play07:30

I confirm that it will work so then we

play07:32

get the transcript back that's actually

play07:33

part of the prompt where after the call

play07:34

is complete can you please give me a

play07:36

summary of what happened so yeah check

play07:37

out the prompt check out the schema I've

play07:39

done all the hard work for you here to

play07:40

make sure this is outputting the kind of

play07:42

responses you want so yeah go have fun

play07:44

with it use it for your personal use

play07:45

start to put in your own Solutions

play07:46

whatever you want to do if you've

play07:47

enjoyed the video please hit down below

play07:49

leave a like and also subscribe to the

play07:50

channel if you want more content like

play07:51

this if you're interested in learning

play07:52

about other AI voice use cases like

play07:54

setting up a customer support phone line

play07:56

for your business or for your client's

play07:57

business you can check out my other

play07:58

video here that's very short and sweet

play08:00

with a custom knowledge base but aside

play08:02

from that guys that's all for the video

play08:03

thank you so much for watching and I'll

play08:04

will see you in the next one

Rate This
β˜…
β˜…
β˜…
β˜…
β˜…

5.0 / 5 (0 votes)

Related Tags
AI CallingAutomationChatbotPersonal AssistantTask ManagementAPI IntegrationEfficiencyBusiness CommunicationTech TutorialProductivity Tool