29-June_22__202 - SCAI004
Sloup
1 follower
1 Video View·Sep 8, 2025
02-Live Call Recordings\01-Live Recordings\29-June_22__2024 - SCAI 004
I think it's specific to my account.
Okay, so you went to chat GPT and then what was the issue?
I can't find the plugins.
Was that?
So yeah, let me show you.
Yeah, I need to re-record that section.
But they actually changed the name for recording.
Yes, sir.
They changed the name for GPTs.
They changed the name for plugins to GPTs.
There we go.
So problem solved.
And then I asked chat GPT, I said,
is there a difference between a plugin and a GPT?
And it's like, yeah, the plugin is and a GPT is.
You solved my problem.
No, it's if you just click on explore GPTs, it's all here.
This was an update that they did last year.
Yeah, yeah, so it's all here.
You can create your own custom GPT too.
Did you know that?
I did, but I have no idea what that would entail.
So I will wait to try.
Okay, well, I mean, there's only three of us here.
So I mean, tell me more about your business or what you do.
Maybe we can do one now.
Sure, I'm a financial advisor,
both on the insurance side and also the investments side.
So I was looking at this course to get into
leveraging marketing.
And really actually I'm leaning towards
focusing on the property and casualty insurance vertical
to be an AI agency because their regulations
are way less intense relative to someone
who's working with investments.
I would rather focus on the agency side
than my own business side,
if that answers your question in a very long way.
Yeah, so you want to start an AI agency
on top of your financial stuff that you're doing?
Instead of.
Oh, instead of, okay, got it.
Jamie, what's your background?
Oh, you're muted.
Okay.
Okay, so basically I'm in real estate
and I got to, I was able to experiment a bit
with automation and integration
because I've been basically, you know,
trying to streamline a lot of my tasks.
So I have a little bit of like
where you think automated videos
and images with free to make and badly
and trying just basically experimenting
a lot of things with GHL
and putting all of these automations together.
But right now what I'm trying to do is,
so I'm not from the US,
I'm from the Philippines actually
and we close a lot of sales through messenger chats
and that sort of thing.
And it's a really big pain,
especially when you're being,
when you're dealing with a lot of inquiries.
So I'm experimenting on voice flow
after going through your course
and I realized that it's really tough
to train the knowledge base, right?
Like sometimes it comes out really weird
and I think I just wanted to understand
well, what's your process?
That's one of the targets.
What was the guidelines that you need to be aware of
when you're trying to train a knowledge base?
And also, and personally,
I'm trying to see how much of the sales process
can it take up?
Because it's normal for us to be able to close sales
only through messenger chats
if we worded the right way.
So I'm trying to see how much of the process
can be replicated with voice flow and that sort of thing.
Yeah, yeah, with AI.
Yeah, absolutely.
So do you, are you using it as like appointment setter?
So they would go onto your messenger
and book an appointment and then you close them?
Yeah, okay, yeah.
I know that process well.
Hey Art, thanks for coming.
So I guess there's a,
I mean, I think we can take this call today
in a few directions.
I can do like a custom AI assistant in chat GPT.
We can do the messenger.
I mean, and I think in the chats,
a lot of people voted for WhatsApp and image generation.
So there's so many topics that we can cover.
So if anyone has like specific questions,
let me know in the chat.
I think I'll go with the first one,
that question, just to show you.
By the way, this is like interactive session.
So if anyone has any questions that you want me to cover,
like any tutorial you want me to do live,
just let me know in the chat.
Hi, John, thanks for joining us.
And so GPTs, just to cover this portion,
GPTs is basically a way for you to have an extension
to chat GPT.
You can almost think of this as like the Apple store
back in the days.
And then these are all the apps that you can have
on your iPhone or Android.
So like, for example, if I search for Expedia,
like if I went to the Expedia website,
I would have to look up my flights or whatever,
where dates and flights and whatever.
But here I can just get a travel GPT
and then it's pulling that information from Expedia.
I can just talk to it and say like,
I wanna go to Paris on July 1st to July 5th.
Like what's the prices you offer?
Stuff like that, right?
So currently there's two ways
that you can monetize these GPTs.
One is you can put it onto the GPT store
and some of these GPTs are paid,
which means that OpenAI will give you a percentage
of all the sales that come through.
And the other way that you can make money from GPTs
is to make it for specific companies,
like specific clients, basically.
I have a friend who does this
and he charges $2,500 per GPT.
So it could get pretty lucrative.
So let me see here.
Where's the menu?
And it's actually not very hard
to create your own GPT here.
They change things around again.
Whoops, where is my...
Hang on, let me find my...
They change things around all the time here.
Oh, yeah, here we go, sorry.
The button used to be at the bottom
and now it's at the top.
The UI always changes.
So you can go to my GPTs here
to see all the GPTs that you have
and you can always create a new one right here.
Okay, and so there is...
So I created one here as a demo.
Called Helena's Bake Shop.
And it's actually part of my YouTube tutorial too.
I'll show you how this is done
and then we can create a custom one here.
So basically, I made an imaginary bakery
called Helena's Bake Shop.
And here is a description.
You're an FAQ bot for Helena's Bake Shop
here to answer any inquiries.
And then this is the exact instructions that I gave it.
Here's the conversation starters.
And then you can upload your own custom knowledge base here
because if this was an actual store,
they wouldn't know my opening hours.
They wouldn't know my refund policy, stuff like that.
So I can add in all these custom information
just about my business
and essentially create a custom version of ChatJPT.
That's why it's called CustomJPT
for my bake shop.
So just to walk you through the capabilities of GPTs,
you can have your name, description,
the instructions here, any conversation starters.
So this is like, you see the four options here
that pops up.
And then that's like the four options that pops up.
You can add in custom knowledge
and then you can give it specific instructions here.
All right.
So you can give it the web browsing capability
so that they can browse the web.
You can allow it to generate image
if you have this checked.
And then finally,
if you want it to be able to read code or data,
you just select this.
But for the Helena Bake Shop,
GPT just doesn't really make sense
for it to start generating images.
And then the last portion here
is that we can add a action here.
And then this is where you can actually connect it
to APIs.
The code is harder to write
because you have to write the API JSON code from scratch.
Whereas like make.com or Zapier,
it writes it for you automatically.
You just like drag and drop the modules that you want.
So here you do have to write the code yourself
for API connection.
But let's say,
a customer put in their name and their email
and I want to store that in my MailChimp account
or any other CRM,
I can actually make that connection
and then actually use GPT as a way to collect leads.
And this is great because if you,
right now, anyone can publish their GPT
onto the OpenAI store.
And just imagine like,
for example, Zach, for you,
you were saying you wanted to do financial data, right?
So let's say you created a GPT with everything that,
sorry, that name is just coming to my head.
The really famous investor.
Berkshire Hathaway, yeah.
Yeah, what's his name though?
It's just like right there in my head.
Now, I don't have it either.
I know exactly.
Warren Buffett, Warren Buffett.
Let's say you-
Warren Buffett, yeah.
Let's say in the knowledge,
you upload all of the documents
that Warren Buffett has ever written or said,
upload all the Berkshire Hathaway annual letters
and then you can have,
actually, we can even create this GPT now
and you can upload all the documents
so people will be able to see your live GPT in the App Store.
They're like, oh, I really want to know
what Warren Buffett would do in this case
and then you can say,
and then you can even add an action
to then get them to sign up to your newsletter,
financial newsletter or something, right?
So you can actually use these GPTs
as a way to collect leads.
Now, before we build it,
I just want to actually also,
basically trying to kill two birds with one stone
with what Jamie has asked earlier
about having this on Facebook as well.
Now, you can, these GPTs,
it's not possible to use them off of chat GPT, okay?
It's kind of like if you had an app on your phone,
you can't use it off of your iPhone.
You need an iPhone to use it.
So GPTs lives on chat GPT.
And if you want to actually use your custom GPT
off of chat GPT,
you have to create something called OpenAI Assistant, okay?
So then you go to this,
I just Google OpenAI Assistant,
then you log in
and then it actually just takes you to this link here
where you can create your own assistant.
I think it just platform.openai,
I don't want the docs here.
It's giving me the technical docs.
Yeah, okay.
So if you go to platform.openai.com,
it takes you to this page here.
This is where you can create your API keys
and your assistant.
You want to go to the second one called Assistant
and you can see from the options here,
it's quite similar to what we just saw
with the custom GPT here,
like name instructions you want to give it,
which model you want it to use, right?
And these models all have different prices
and you can add in your search files,
a code interpreter,
if you want to turn on this function.
This is the API policy you can call.
Am I losing anyone?
Does this make sense so far?
I have a couple of questions.
Yeah.
Can I ask?
Okay.
So when you're talking about
uploading files for the knowledge base,
is the code interpreter and these other functions
part of that knowledge base that you're,
is that what you're turning on and off
or what's the connection between those toggles?
Code interpreter is to write and run code.
So if you're just...
I understand that.
Okay.
So I'm just wondering,
is that relative to whatever you put in the knowledge base?
No, that's file search.
So if I want to, let's say for a Helen's Bakeshop,
upload a bunch of options for what I offer on my bakery,
I would just turn on file search
and then upload all the files here.
I don't need code interpreter for that specific.
But I'm just saying, when do you use...
So code interpreter allows you to upload code
to the knowledge base
and then it can interpret that data.
Yeah.
It's for running...
Okay.
I'm just clarifying that.
And then is this something,
this is like a really simple workaround
where maybe you don't want to do a full-on bot
on someone's website,
but rather have someone just click a link
to the URL for the GPT?
Is that some kind of an easy workaround?
Or am I thinking about this incorrectly?
Yeah.
So if you have a custom GPT,
they have to be on the plus version of OpenAI
to be able to access it.
Ah, okay.
So they have to have an OpenAI account
to be able to use it.
It's great because it gives you that extra...
Like I said, there's two ways
that I see people make money with custom GPTs, right?
One is they actually make a custom one for a client.
Price probably range from, you know,
a thousand on the low end, 2,500-ish.
That's like kind of the range I'm seeing at the moment.
And then the second way
that you can make money from GPT
is to actually just put it by yourself
on the OpenAI store.
Like yours can show up here.
Like anyone can put one on here right now.
Like it's the wild, wild, wild west at the moment
when there's not too many rules and regulations yet.
And if you get popular,
you know, you can even land on the feature section
if a lot of people are downloading your custom GPT here.
And the way that, you know,
HubSpot is making money with this GPT is they're...
I mean, I don't know how they built it.
So I can't say for sure,
but I'm guessing they're probably trying
to collect your information
that they can then store in their CRM system.
So it's a lead generation system for them.
It's a brand recognition for them
because they have their logo here.
And then just similar to the App Store,
some of these GPTs are paid, some are free.
And so whenever there is a payment,
OpenAI takes a percentage
and then the creator takes a percentage of that.
So that's kind of how these GPTs
are currently being monetized.
Essentially OpenAI kind of went
after the Apple model,
like how Apple started the apps.
They basically copied that business model from Apple.
Does that make sense?
Did I lose anyone now?
Yeah, Art, yeah.
Yeah, so my question is with the custom GPTs,
this is sort of like a store for people
who have access to ChatGPT to explore.
How about people who are not on ChatGPT?
Would I have to direct them there to get access?
Or how does it work?
I mean, I'm talking about the general public
to actually see this.
So you have to be on the plus version of ChatGPT.
So I think it's like $20 a month or something
to be able to access the custom GPTs.
So if they're not on the plus version,
they won't see this whole section.
Was that the question?
Yeah, that's what I'm trying to get into.
Like for the client's consumers,
if they wanted to have access to this ChatGPT,
they would have to have a subscription to do that.
Or like I was saying, you would create an assistant.
So assistant would use API usage.
And so basically, you can create a custom assistant.
And then I'll show you.
And then you put in your credit card.
So whenever this assistant generates output,
you will be charged according to the model that you're using.
And so you will pay for it, basically.
So let's say, for example, I select GPT 4.0 model.
And currently, the pricing is $5 for 1 million tokens of input.
And then $15 for 1 million tokens of output.
So I can create this assistant,
like basically copy over whatever I put in the custom GPT,
create this assistant here.
And then I can use it off of ChatGPT.
But I need to, you need to pay for,
it's not like a set monthly fee.
It's just based on how much it is used.
So it is possible to take it off of ChatGPT.
It's no longer called a custom GPT anymore.
It's called an assistant.
And how do we translate tokens into something?
So it's a very, I'm glad you asked.
It's a very complicated math system, actually.
So I'll just say like, hi, my name is Helena.
This is 20 characters, as you can see, and six tokens.
So OpenAI can actually help you calculate this.
It's generally four to five characters equals one token.
But if you look on YouTube,
there is, it's like a crazy math equation.
Someone did a YouTube video on it.
I watched it and I was like,
I'll just stick to four to five as my guideline.
But so generally you just like go to word count
and like Google Doc or something,
look up the number of characters,
divide by four, divide by five,
and then you'll get, you'll be within the right ballpark.
Does that make sense?
So one million tokens is roughly four million characters.
So it's a lot, roughly.
Like, you know, like it's a very complicated way
how they calculated it.
Honestly, it went over my head.
It was like PhD level math.
About it.
But there is like a 30 minute YouTube tutorial on that,
on the calculation.
But honestly, I don't think that really adds
any additional value for us.
I think just using the rough rule of thumb,
which is roughly four to five characters,
it's enough for most people's understanding
and usage general.
That make sense?
Okay.
Anyone have any other questions regarding that?
And then so Jamie, so like,
for what you said you wanted to do,
you can link like Assistant to like Facebook Messenger,
for example, or like WhatsApp and other tools.
So if you want to use your GPT,
or this can be linked onto your website too.
So basically, if you want to use your custom GPTs
off of OpenAI, you actually want to create an Assistant,
not a custom GPT.
But let's just create one
so that you guys can see the whole process.
This is actually a GPT helper.
So let's say, using the example we did earlier,
I want to create a custom GPT for financial advice.
And I want to link all the best practices,
and lessons, strategies that Warren talked about.
Okay.
So you can actually just conversationally create
this GPT in the Create section.
And then it's automatically going to fill in.
I'm not sure who was coughing.
If you guys don't have a question,
you can mute yourself, that would be great.
So we can see here, I did this,
and then it's starting to create this GPT for me.
So context, this GPT provides financial advice
focusing on the best practice,
investment strategies used by Warren Buffett.
Okay. Something went wrong.
Yes.
Oh, something went wrong.
Okay. Sometimes it does time out.
We're back.
I want to create a GPT for people looking for financial advice.
So you can use the Create tab if you want the AI
to create their GPT automatically.
If you go to configure, you would do everything manually.
I'll show you.
I'll show you guys what it looks like.
Okay.
And then when you see this icon here,
it's actually just starting to create that GPT.
So it's creating the GPT,
it's going out and grabbing data from the internet.
Is that how it's doing its thing?
Nope.
What it's doing, it's filling out this section for us.
It doesn't do-
I see.
Yeah. So like-
It doesn't do the knowledge base for you.
No. No, it does not.
Okay.
Like if I want to chat, like, and I start,
go back to the GPTs, create.
If I go to configure, see how that's all empty.
If I'm chatting,
if I'm chatting with chat GPT,
like in the Create section,
it just automatically fills out the name, description,
prompt, and conversation starters for me.
Okay.
You still have to go in and add in your own knowledge
and any actions you need.
So you can see it's actually pretty easy
to create your own custom GPTs.
And then you can see here,
when you're ready to launch it,
you can test it here, you know?
So I said, what are Buffett's key principles
for successful investing?
Oh my God.
Got the hiccups now.
And there we go.
And it gives me the answers here, right?
So like I can make a Helena GPT
if let's say I upload all my YouTube transcripts,
everything on my course,
and I'm like, you know,
I want you to just learn everything about me.
And then people can ask me questions, right?
So that's,
and then I can publish it now here in the GPT store
and then save it.
I'm not gonna publish this.
And then another thing you could,
so the only hard part about creating these GPTs
is adding in custom actions.
So like, this is where you had to like,
add in the actual JSON code for specific actions.
Like if I wanted the weather,
you just go to example and then click on JSON.
JSON's probably the type of code
usually used to get data.
Although now with web,
now that these GPTs have web browsing,
you generally don't need to use this
because you could just say in your,
let's say I wanted the,
where GPT, like, sorry,
where these actions were used before a lot
was because these GPTs didn't have live web browsing.
Because like, until like two weeks ago,
OpenAI knew data generally from like six months ago.
And actually if you don't use the latest model,
which is GPT 4.0,
like GPT 4, GPT 3.5,
the data is six to 12 months ago.
Like it's not trained on data in real time.
And GPT 4.0, which has just literally been launched,
now has real time data, right?
So like, let's say I wanted to be able to access
like stock prices live.
Before I would need to create an API connection
to like a provider that provides like stock data.
But now I don't even need to do that anymore.
I can just give,
I can just in the instruction say,
go to this website
and then look up the stock price from here.
Like go to Yahoo Finance
and then look up the stock price from here, right?
So that's what actions were used for a lot
because GPT didn't have the latest data.
So you needed to pull in an API to access a lot of data.
Like weather as well.
GPT doesn't know the weather today,
but now with 4.0,
it can actually access like weather.com
and get the weather today now.
So there's less use cases for the action
than previously, than what there previously was.
Because now GPT does have the ability to browse the web.
Yeah, question?
Yeah, sorry.
So going back to my very first question,
which was, it is web browsing to get the answers
that you just asked it
because you have web browsing click.
And so one can,
and instead of uploading a bunch of files,
if we could just create action by connecting to websites
that already have that data,
URLs that have that data available.
So it'll just pull.
Yeah.
Okay, cool.
And then you upload your own,
if you had his letters for saying
the letters aren't available out on the internet,
then you would upload those letters.
Yes, yes, like, yes, exactly.
Exactly because OpenAI has only been trained on data
that's publicly available on the web.
But if you have specific data,
you know, regarding your business or industry
that's not on the web, it won't know that.
Let's say if I have like a SOP
for how to do customer refunds or something,
you know, that is not publicly available,
that's where I would upload it.
And then like, let's say I made a custom GPT
for training my new employees
and none of my SOPs are publicly available
on the internet, obviously.
So that's where I would upload all my SOPs.
And then like, if I have a new employee
that was onboarding,
they could chat with the GPT to get the information
they need to start the new job, right?
That's like another way that you can use this, you know?
And so I have a question about what you just said.
This, by putting this into a GPT,
it's somewhat public.
So something like, you know,
the SOPs may not be something you want to use a GPT for.
You want to do a private bot
that holds the data internally for compliance reasons?
Yes, yeah, depends on-
Okay, cool.
Yeah, depends on the-
Okay, I'm just checking.
Yeah, so anything you upload here
could be used for training data.
Great point to note.
But it depends on the company.
Like, you know, some company may not really care.
And so like, if it was like a bakery,
like, yeah, you know my SOP,
it doesn't really matter.
And this would be a lower cost solution
than say an internal chatbot, is that right?
Exactly, exactly, exactly.
Okay, perfect, perfect.
So they need that kind of security,
you know, everything raises the price, right?
So it's, yeah.
But yes, one thing to be aware of
is that anything you put in here
could be used as additional training data
for OpenAI.
And if security is a big concern,
then this may not be the solution for that.
But there's definitely other solutions.
Like OpenAI, I think like 10%, I want to say,
but don't quote me on that,
is actually owned by Microsoft.
And Microsoft actually has a suite of enterprise tools
to build like chatbots,
build these knowledge repositories and stuff.
It is on the pricier side,
but it's meant for enterprise.
And it uses OpenAI technology,
but the data is secured
and is not used for training data.
So that's kind of like the enterprise solution.
Microsoft is already on the ball with that one.
Yeah, yeah, but great question.
I get asked that question a lot
in terms of security.
Yeah, so generally if it's like,
depending on the size of the company and their budget,
because Microsoft is expensive.
Microsoft solutions for enterprises is expensive,
but they have all the compliance
that large enterprises are looking for.
You know, yeah.
Any other questions?
Just trying to sum up this.
So generally anything that you upload
to CloudGBT will be used publicly
or will be used for the training CloudGBT.
If you want to keep it private,
then you have to leverage some of the enterprise tools
that you just mentioned or build your own
and then use the API to call CloudGBT
to do whatever needs to be done.
Yeah, so if you're on the Plus plan,
anything you enter into GPT can be used.
Like every day when I...
Yeah, just one second.
Okay, I'm just gonna mute.
Yeah, so to answer your question,
if you're on the Plus plan, yes.
Like anything I'm currently entering into ChatGBT
can then be used as their training data.
And I think they...
So OpenAI just came out with a team option as well.
So for team, you can...
Yeah, team data excluded from training by default.
So if you're on the team plan,
you need minimum of two users.
Each user is $25 per month into $20 per month.
So if you're on the team plan,
whatever you enter into OpenAI
is actually not used as training data.
I'm on the Plus.
So anything I'm using can be used as training data,
but I'm not putting in anything crazy.
So I really don't care.
And for most individual users, it really doesn't matter.
But if privacy is a huge concern,
then upgrade to the team option.
Yeah, it's just like...
We're working with other clients that may require that.
I just wanted to understand what the boundaries are
so that I can know that,
you know, they would have to go
to an enterprise rescue center to upgrade themselves.
Yeah, so you would just upgrade to the team option
for your data to not be used.
So it starts at $50 per month.
You need at least two users, $25 per user per month
to be able to do that.
Can I kind of see if I understand the mapping out?
Because compliance is going to be a big issue
if we're dealing with clients, right?
To varying degrees, right?
So at the very basic, nobody...
We don't care.
It's the Plus version.
Little more security would be the team version.
More security than that would be like
using an internal AI agent on the website.
So that's kind of the next version.
Then enterprise would be the most expensive hardcore.
Is that fair?
Yeah, so Azure is from Microsoft.
And because Microsoft owns a percentage of open AI,
they have native thinking.
But basically you're then wanting to find
enterprise level solutions.
At that point, you're probably...
Like where I see your role is probably just being
kind of like the business development person
or like the person closing the deal.
And then you probably hire a developer
who specializes in this to actually do the work
of linking it.
Gotcha, okay.
Yeah, yeah.
That's because then that frees up your time
because if you're building and then like,
well, your time is better used to get another customer
and then just pay a developer work.
That's how you can match how much you earn, right?
So what I'm trying to do is I understand that.
I'm trying to figure out if I go into a client
and he's like, yeah, I want more security than Teams,
but less security than Azure.
Then I do an AI agent for their website specifically, right?
That would be kind of in the middle
in terms of security?
Yeah.
Okay, perfect.
That's what I was going.
Thank you.
Because Azure has a lot of cybersecurity stuff built in too
that's actually beyond the scope
of just making sure that the generation
that you get from open AI is secured.
Like for example, they have like, you don't get hacked.
Like your whole server don't get hacked
and stuff like that.
Like it does pretty much everything
and it's basically impossible
to know all of these platforms.
It's like where I see kind of your role
is just understanding it.
Then when you actually sign the deal,
then hire a developer to actually do the work.
I wouldn't recommend trying to learn all of this
because at that point,
you've just found yourself a job as a developer
rather than running a business, right?
We're starting an AI business at that point, right?
So that's a route that I would take with that, yeah.
So another thing you should know
is that the API data is actually not used for training,
which is great.
So if I use assistant by default,
the stuff that is entered in and coming out
is actually not used by open AI for training purposes.
That's another thing that you should note,
which is great.
And you can always connect
if there's a cheaper solution.
Like Azure is not cheap,
but it's made for enterprise.
But something like Go High Level can be connected
and you can always get HIPAA compliancy
on Go High Level as well.
I think it's a 497 extra per month.
You can get HIPAA compliancy
and you can even use it for medical clients as well.
Not sure that's a route that you guys wanna take on,
but it is possible as well.
Yeah, so that's basically the overview.
And I was showing you here,
it's the same process to create an assistant,
name, give it the instructions.
So when you're happy with this GPT,
you can just copy over these instructions here,
select your model,
add any custom documentations you want.
And I think it automatically saves.
So you don't have to press save.
And then you can try it out here
to make sure you're getting the results that you want.
And then when it's ready,
you need this assistant ID here
to then use it off of the platform.
Yes, Art.
So the instructions that you just mentioned,
that's the points that you have to call specific use case
that you're doing that, the instructions there?
Yeah, yeah, this is your prompt.
Is it?
Your prompt, yeah.
Yeah, like I'm just gonna use in this example,
like the Warren Buffett example again,
like this is my prompt,
like this GPT will serve as a financial advisor.
The role is to act as a virtual financial advisor
in imparting Warren Buffett's investment wisdom
to the user, the constraints,
avoid providing specific stock recommendations,
always advise the user to consult a professional.
Financial advisor do not offer advice
on short-term trading and speculations.
So you guys already know that like this is a prompt, right?
And then it created it actually automatically
when we were in the create section.
So if I wanted to use this off of the platform,
I generally just use make.com
and I can connect this assistant with make.
So it depends on what I want my,
so like if I went into my make account
and created a new scenario,
so I need something to like trigger this,
like trigger this GPT.
Yeah, okay, so I have this saved.
Okay, I'm trying to think.
So let's, okay, let's do a Google Sheet.
I'm just trying to think of the easiest way
to show you this.
And let's just say question and then answer, right?
So question for Warren and then answer for Warren.
Of course, I'm just showing you the mechanics.
I know that you're probably never going
to need a spreadsheet,
but the general process is the same.
So like if I did Google Sheet
and I wanted to watch, yes, watch new rows.
So when a new row, new question is entered,
then that is where I want this whole automation
to be triggered.
And I'll call this Warren Buffett questions,
the test, okay, for myself here.
And then I have to now find this spreadsheet.
Yeah, Warren Buffett test sheet.
The sheet name is the first one.
That's the only one we have.
Head table contains headers, yes.
And I wanted to do one row,
like do one row at a time.
Going to press okay.
And then the next one, I'm going to search for OpenAI.
And I'm going to search for message and assistant.
So I've already connected this to my OpenAI account
because I made it in my account.
So I should already see from this dropdown list,
like the Warren, yeah, see here,
the Warren Buffett investment advisor that we just made.
And then the question that we want to ask here
is like we want to input A
because you see how in column A,
like the header is question.
So basically what this automation is saying
is whenever there is a new row being entered,
the question that we want to ask
this particular system is the cell,
the new cell in the new row.
And then I'm going to press okay.
Does that make sense to everybody?
Make sense?
Any questions?
Yeah, give me a yes in the chat
if that so far makes sense.
Okay, do they know?
Yes, okay.
I'm unclear what you're trying to do
with question A and question and answer.
Are you going to input a bunch of questions and answers
and where is that?
Yeah, I'm just showing you like the easiest way to show you,
like you can use assistant,
how to deploy the assistant off of chat GPT.
So like how to use it off of the path,
off of open AI,
like this could be a fork, chatbot, messenger, et cetera.
But so like I'm going to enter in a question here
and then it's going to go into open AI into this.
Okay, it's an instruction
and you just happen to put it into a spreadsheet.
Yeah, yeah, just to, yeah,
because it was the fastest and free way to,
it's the fastest and the most, it's free to do it.
So without any additional tools,
but just to show you like you can use a system off,
like on any platform basically.
And then now I need the response from the assistant
to go into this cell here.
So I need to start another module.
I had to select a Google sheet again
and let's see if I can just update a cell,
update cell, that's what I need.
And spreadsheet, select from list.
I'm going to select the same spreadsheet that I had before.
Yeah, Warren Buffett.
And the sheet name is going to be one.
The cell is going to be the answer cell.
And the value of the cell will now have to be,
this one, the answer that we just got from the assistant.
Okay, so let's try this out
and see if I hopefully did not make any mistakes.
I'm going to copy this question here,
which is what are Warren Buffett's key principles
for successful investing?
Go back to make, let's run this.
And then we're doing this live.
So let's see if it works.
The question, what are the pros and cons
of doing this directly for open AI,
like doing that system there and then using it?
It just, you need someone to deploy the assistant.
So basically you make it in the open AI assistant platform here,
but you still need to connect it to something to deploy it.
It could be custom code.
It could be a no-code tool like Zapier or Make as well.
But you still, like essentially you're just,
you've created an API connection to a custom GPT.
And then you can connect it to whatever you want
to connect it to, right?
Missing value in parameters cell.
Okay, so let's first, okay.
So you can see here on the top,
this helps you to diagnose what doesn't go right.
Let's first see that the content is correct in text.
Okay, so it's actually giving us the right answer to our question.
It's just not posting it into the answer cell.
Now, so I would, in this case, put B3.
Okay, instead of making a dynamic,
I just like test it to make sure it works manually.
Like it's a little bit of like troubleshooting sometimes,
like to get it working properly.
So let's test that again.
And then this should be okay.
And then we go back here.
You can see the answer is now posted, right?
And then, so I see that.
Okay, so it was just this cell value
that wasn't being pulled in properly.
Okay, so I actually do this.
Okay, so now this should work.
So it's going to column B
because I always want it to be in the column B.
And then pull in the row number from that.
So I just pulled in the wrong cell.
And then I can save this again.
And then I can put in a new question.
And I'll have the new answer here.
Obviously, I know you probably don't always need it
in spreadsheet, but the principle is the exact same.
You just replace the node here
with whatever you need it to be.
Like, again, it could be anything you want,
Messenger, any chatbot, Facebook, Instagram,
WhatsApp, any of them, right?
There's literally infinite number of use cases.
But at the end of the day, the principle stays the same.
And if you understand the principle,
then you can create anything you really want to.
It's kind of like, if I know the principles of adding,
it doesn't really matter what numbers you throw at me.
I don't need to hard memorize, like, you know,
two plus five equals seven, because I know how to, like, add.
So you can throw any two numbers at me
and I know how to add, right?
So that's kind of the same principle
that you want to keep in mind.
We're just running up on the hour.
Does that make sense?
Any last minute questions
before we end today's session here?
Learn something new, guys?
Yeah, just a quick question.
You can just show us how to integrate with WhatsApp.
I don't have time with four minutes,
but I can do it next week, yeah.
That's definitely more than a four-minute conversation.
I can definitely do it.
Just if it was a quick question, I would definitely do it.
But that's, like, 30 minutes to one hour
to have that properly explained.
Let me see if I have maybe something saved.
Let's see if I have...
Let's see if I saved it.
Oh, no, I made it in here.
Okay, it doesn't look like I can find it.
And I think I made that in a client's account.
But I can show you this, which is for Instagram,
which is in my account, which gives you the same idea.
This is to get a custom response from Instagram.
As you can see, it's pretty similar.
Like, the message comes through from Instagram,
goes into this webhook.
It's the exact same process for WhatsApp.
Goes and messages an assistant,
in this case, Helena's Bake Shop.
And then it converts it back into a JSON structure
and then sends it back into a webhook
and then displays it to the user.
You just need this extra JSON step
when you have to convert it back into a message
so that, like, Facebook and Instagram
can actually read that message.
We didn't need to convert it here
because it's fine to just display the message as it is.
But if you need something like Facebook or Instagram
to read that message,
you need to put it in a language that they can understand.
So it's the exact same step.
It just has an extra step for the JSON.
But I can go, I can, like,
it's basically the same thing that I just showed you.
But I can go into this next week
if you want to build it again from scratch.
But I hope...
Yeah, that would be good.
Yeah, I think that's helpful.
It's just understanding how the webhook gets called
from WhatsApp.
That's, that's...
Yeah.
Yeah, I can do that next week for sure.
Okay.
All right, guys.
Well, I want to thank everyone for coming
and I hope you learned a lot from today's training.
And I look forward to seeing you guys next week.
You're going to post the recording in the classroom?
Is that where it's going to be?
Okay, cool.
Thank you.
Yep.
Thanks, Zach.
Okay.
See you guys next week.
