Building a Custom GPT Lead Generation Chatbot

In this video, I'll be showing you step by step how to create an insanely powerful lead generation custom GPT with multiple custom actions and add it to any website. If you're watching this, you've probably already created some awesome GPTs within the ChatGPT builder, but you want to break them out of the ChatGPT site and start using them in the real world. Or you want to learn more about how to add powerful custom actions to your GPTs to make them more useful for your use case. Unfortunately, a lot of people still don't understand what the new assistance API is, and even I was confused by it at first, but knowing what it is and how to build your own custom GPTs with it puts you in the top 1% of people who are chasing down this opportunity. So by the end of this video, you'll not only know how to build your own custom GPTs using the assistance API and put them on your websites, but you'll also know how to plan, create and implement insanely powerful custom actions into your GPTs to create real value for your businesses or for your clients. And at the end of the video, I'll be giving away everything that I'm building today as a template so that you can steal this insanely powerful lead generation chatbot and start using it in your own business or selling it yourself. If you're new to the channel and don't know who I am, my name is Liam Otley and I run my own six-figure development company called Morningside AI. And I also run the largest community of AI agency owners in the world where my students learn how to make money by selling AI solutions to businesses Just like the lead generation chatbot that you're about to build. Let's get into it. To kick things off, I'm actually going to be showing you the finished product here on screen so that you get an idea of where we're going with this and how much value we are creating when we do this right. So on screen is a template website that I've downloaded. Just refreshing the page so you can see that it's live. If I click on the widget in the right here, up pops the chat window and I can start talking to it. Hi, I'm Lucy, Smith's solar assistant. I can help you with any questions on solar for your home. I can also calculate the potential solar savings for your property. How can I help you today? That gives you an idea of the two functionalities of this chatbot. So firstly, we can ask questions to the knowledge base. What is the cheapest type of panel? This will be querying our knowledge base document, which you guys will get to see in a second. The general FAQs and information brochure for anything related to solar for single family homes. Hugere we have the answer. The cheapest type of solar panel is polycrystalline solar panel. Great. So that shows the knowledge base is working, but more importantly, here's the feature. Hugelp me calculate my solar. Solar savings. Hugelp me calculate my solar savings please. The assistant is going to pick up on this and going to realize that we need two inputs which is going to be their home address and their monthly electricity bill in USD. So for this I'm picking someone's lucky address in the United States. So let's grab this one here, click on this, sorry to whoever's home this is, broadcasting your address out, copy this, head back and if I just put in just the address here it should ask me for the monthly bill as well because it needs the two inputs to calculate this. Great. Thanks for sharing your address. To proceed, I'll also need your average monthly electricity bill. So let's say I'm spending $250 per month. Now it's going to take that address, it's going to take that monthly cost of power, and it's going to call the Google Solar API, bring it back, and tell us exactly what kind of savings we can expect for our property if we choose to convert to solar. There we have it. It's given us a full breakdown of our options for solar based off allocation. I'll break this down a little bit more later. It's really, really complex stuff that's going on behind the scenes here. But essentially, it's giving us information like... our payback period and the 20-year savings we're going to experience, any federal incentives that are associated with our location. But that's not even all this thing does. At the end, it asks for a lead capture. So it's asking me for my phone number and my name. And then if I provide that, I go Liam Otley, and I give it a phone number, phone number, and I give that to them. Then what this is going to do is take that information. and it's going to also take my address and it's going to push that all to a crm which i'm going to show you how to set up later and then we have a thank you your details have been passed to our team someone will be in touch with you soon so that's the breakdown and a demo of the bot but before we get into the building i need to do a little bit of explanation so that you're fully clear on the difference between creating gpts as you are on chat gpt and creating your own cool ones within the chat gpt website and the kind of thing that we're building over here which is using the assistance api these are your two different ways of creating gpts now on the chat gpt builder they are locked inside of that With the interface, you can't take what you create on the ChatGPT site when you build these things and then go and slap it on your website or put that on WhatsApp or put it wherever you want. You are locked into that and it's intended for you to just use it in there as a handy little kind of personal tool. I've said that GPTs in this way are like consumer toys. They're focused for the mass market and consumers to engage and have fun with and build these GPTs. But the other side of this is using the assistance API. I've been over this recently in my other videos, but it's important that you understand the difference. and that most of the value is going to be unlocked through this assistance api because this allows businesses and anyone really to take that same gpt functionality that you create there which is really cool but then take that and put it wherever you want integrate it into any application any website any voice assistant whatever you want so what i'm teaching you in this video is how to use that side of things the assistance api which is far more powerful and is going to create far more value in our world and really be a much more valuable skill than being able to make a little fun gbts on the chat gbt website then we need to get into a little bit of a deeper breakdown of what we're building and how we're going to build it so if you haven't figured out already we're making a lead generation chatbot for solar companies that's going to be put on their website which is intended to help leads and people who are interacting with the website to get you answers to their questions. Firstly, if they have any questions related to, hey, what solar panels are best? What financing options? What's this? What's that? Anything that they want to ask as if they were talking to a solar expert, we're going to allow them to ask that to the chatbot, which can respond from a knowledge base using a document that we're provided containing this kind of information. Secondly, the chatbot will also allow them to calculate their potential solar savings if they were to install a system. And we're going to do that by taking in their address and their monthly power bill and sending that off to Google Solar API. and getting the response back from that. And then finally, once I've got that back from the calculator tool, we're going to ask them for their lead information so we can capture the lead and follow them up. We can break it down into my three ingredients concept for creating GBTs, which is prompting, knowledge, and actions. The prompting is going to tell it that it is an assistant for Smith Solar, that it can answer questions from a knowledge base regarding solar questions, and it can also provide calculations, and it can also capture the lead at the end of it. So that's all the information that we're stuffing into the prompt. For the knowledge component, we are uploading a single document containing information, a ton of information on solar for single-family homes and finally the actions are going to be calculate solar savings using the solar api and then capture the lead and send it to airtape the stack of technology that we're using to create this assistant is firstly the assistance api of course the solar api and geocoding api from google replit to host it and voice flow to be the front end for it okay to get started you need to fork my replit template which is going to be available in the first link in the description you can sign up to my resource hub it's 100 free and there'll be a link in the resource hub for this particular video where you can get access to this replica now when you click on the link it's going to take you to this screen and what you're going to want to do is click on the fork button on the right here and we're going to give it a name solar lead gen chat bot and we can fork it if you give it a second it will boot up in here Then you'll be greeted with this here. For those of you who aren't developers or don't know how to code, this might seem a little bit overwhelming, but don't worry, I've set this up in a way where you don't need to do any any major changes to get the kind of results you're looking for. And any ones that you do, I'll point them out very clearly. And it's just a couple of characters or words that you need to change. So we have our main.py file here, which has our endpoints, the chat endpoint and the start endpoint. Don't worry too much about this. And we have the functions.py file, which has all the specific functionality, such as creating a lead and air table, such as pulling the data from the Solar API. Etc. So all of the fun stuff happens in here then at the bottom we have our create assistant function I have gone pretty heavy on the comments throughout this file So when it comes to modifying it and changing the knowledge or the functions There should be plenty of comments to help you figure out what you need to do But I'll take you through a crash course now because I know many of you want to take this change the knowledge change the function So this is how you can do it So for changing the knowledge, which is the easiest one to do we have the knowledge dot doc X Which is referencing the document on the left here. So if you want to change the document you can delete that and you can upload a new one by clicking this. So you can upload a new file and just make sure that the name of the file is referenced here exactly. If you want to add multiple files, you'll need to copy and paste this multiple times and save them to different files. I recommend that you paste this entire function from the top down into ChatGPT and say, hey, I'd like to add multiple files because it needs to add a couple of different rows here, duplicate that. And then down the bottom here, we need to pass in more than one file ID. So that's for changing the knowledge. For changing the instructions, we have this instructions assistant instructions variable reference here. And what this is actually doing is pulling from up here, we have an import from our prompts file. And on the left here is the prompts file that we're talking about. So the assistant instructions and the prompts.py. If you want to change the instructions for your assistant, this is where you need to do it. So by changing this, you can take a look at how to set this thing up here. But by changing this, you're going to change the prompting and the behavior of the chatbot. And the information in here, as you look at my stuff, you'll see that it's referencing how to use the knowledge. It's referencing how to use the tools and what to do after. So one of the cool things that I added in here was that after the assistant has provided the user with their solar calculations, they should ask for their name and phone number so that one of the team can get in contact with them about installing solar for their home. So that is instructing it to after you've triggered that tool, then do this, which is what you'll see later in the app where it actually asks for the lead capture at the end of the response from the tool. So if you want to change the prompting and instructions, you need to come to the prompts.py file, modify that. If you want to change the knowledge, you need to come to the top of the create assistance function and modify this and modify your documents. And when it comes to modifying the functions, it's definitely the harder part, but I'll try to walk you through it as quickly as possible now and make it as simple as possible. So from a high level, we have in the setup for our system, we have our different functions. So again, this bit here is the knowledge tools. So we see retrieval means that it's adding the knowledge base retrieval tool, which is going to handle the knowledge base up here. But the one below it is a function. So we have our solar panel calculations and then we have the properties and the inputs it provides. So we need the address and we need the monthly bill. And both of them are required. And the same down here. We have another function called create lead, which is going to log the lead in the database. It has a little description and then it's going to need the name, the phone, the address, and all of them are required. Now, you don't need to write these all yourself. You can ask ChatGPT to assist you with this and by pasting in this entire function, it's going to show them exactly how you need to do it. So this is your reference and ChatGPT can help you to modify it. Now when it comes to actually what functions are being used, the name of the function here needs to match the name of the function that you want to call. In this case, I want to call this, which is my sort of full calculations function. Again, if I go down to here, the name of the function is called create underscore lead. And at the top, I have create underscore lead. Now this all comes together. when you go to the main.py file. So over here, if I scroll down to the chat endpoint and I scroll down to this section, you'll see in this part here, if the assistant picks up action required, meaning one of the tools has been triggered based off the natural language description of capture the lead details, calculate the solar potential. If one of those are triggered, it's going to come down here and look up for the different function names. So we have the solar panel calculations and the create lead. When the assistant initiates the action, it's going to pass the name through. and we can go okay what's the tool call that it's called what's the name of that okay it matches the solar panel one great then we want to call from our functions file the solar panel calculations and that takes the address argument and it takes a monthly bill argument you can see this back here it takes an address and a monthly bill and in this case we are just pulling from the tool call we are pulling the address as the first parameter and the monthly bill is the second parameter so Might seem a bit complicated, but again, if you paste this in the chat GPT, it should understand what you're trying to do. Same thing happens down here with the create lead. Again, we needed to have three parameters in here, which is the name, phone, and address. And if the function name that was matched within the tool matches the create lead name here, then it's going to trigger that and pass the arguments in and then output it. That might be a little bit complex for some of you, but I did promise that you would understand how to put these things together and how it all fits. You might need to go back and break that down a couple of times, but that is how it works. The next step for you to set this up is to head over to Google Cloud. So cloud.google.com and you need to come on here and create account. I assume you've already got a Gmail account and you can click on go to console. You might need to do a bit of set up and sort of creating an account on the Google Cloud platform. Once you're in, you should see this. And then what we need to do is come up to the top left. Let's create a new project. I'll call this solar. chat bot for this i'll be using my youtube projects billing account in your case you'll need to set one up just so you're clear on the amount of cost i've been using this a decent amount and i've been charged under a dollar so these things are very cheap to use for the api usage so create your billing account you need to put a card on it so that they're able to verify it and allow you to start using it but once you've set up your billing account and put a card on file you'll be able to select it here and then for the organization i'll just leave it as none so it's going to take a second for it to create my project then i can click select project here so that i jump into it then once it's all set up what you can do is click on the apis and services section here and go to library and then what you want to do is search up geo coding geocoding click on the geocoding api need to enable that when you successfully enable that it should pop up with this api key message so you can copy your api key here which is what we're going to need to use in our application to be able to call the solar api So we can copy this, save it somewhere and go continue to Google Maps. Maybe later for the protecting my API key. Then we'll go back to APIs and services. Let's close this on the side. Then you'll see the solar API appear here. So we can click enable on the solar API. Give that a second. and we can scroll down to make sure we have the geocoding api enabled great we have our google cloud set up we have our api key and we have the solar api and the geocoding api enabled you want to grand bit api key and come back over to our replit and if i scroll to the secret section and click on secrets you should see the google cloud api key section here what you want to do is click the edit button here and then paste in your api key here and click update secret next we need to set up our air table to log all of the leads which will act as our temporary crm and we're you can use obviously any kind of crm that you like i just like using air table because i'm very familiar with the api so in this example we'll be using air table i do have a base here that you'll be able to copy if you want to get access to this crm you can copy and clone it over to your account it'll be in the resource hub as well so there'll be a link to click on that then you should see a page like this and you will see this button here that says copy base so if you copy that as your air table account you'll have this all set up as i've got it done here then we need to do one more thing which is to go on to our developer hub on the right here and you will see all of our personal access tokens so you'll need to click on create token up here on the right give it a name we'll call it lead gen chat bot not the best name and we can go to scopes and we need to add data records read data records right and then we need to set up the access so in this case i'm going to go down to smith solar crm and add that and i can create the token it'll then pop up with your api key which you need to copy and then bring back over to replit again and do the same process so click here edit and then paste it in and save the updated API key. The final step for setting up these keys is to get your OpenAI API key. To create or copy your API key for OpenAI you can come to platform.openai.com slash api dash keys and on the page here you will see a button to create a new secret key. You can click this button it'll give you a new key you can copy that back over to here and do the same process of updating the API key. Now I will mention I had a bunch of people asking in the comments that they couldn't get access to the model that we are using in this particular build. That's because you likely need to put your card on file. What you need to do is go to the settings on the side here, you need to add a card for billing. You're not necessarily going to be billed for it because you do have that free allowance but you need to put a card on file so that they allow you to use these models. Just a little issue I saw in the comments of the last video. The very last bit of setup we need to do here is to get our Airtable base URL. So what you can do is click on the profile icon in the corner, click on developer hub. and then we can click on web api documentation and then you can scroll down and click on your smith solar crm and that will take you to the documentation for your particular base which is super handy if you scroll down to the leads table in this case we want to go to Create leads record and it will give you a URL here So we want to copy this all the way up to the end of the leads word and we've already set this up when I Wrote the code this is already kind of integrated into the function All you need to do is head back and come to this create lead function and replace that entire Bit there I already have mine in there so I didn't change anything But you should replace this URL with your URL with all of that done. We are pretty much ready to boot this thing up Let's just save this file. Make sure we've all saved up. Let's go back and give the main.par file a save as well And we can click this big juicy green run button on the top. Give it a second and it's going to boot up. On the left panel here, you'll see that there is no assistance.json and you'll see a file pop up here in a second. So give it a sec. And then we have the assistance.json file pop up. The way that I've set this up is that the create assistance function will run at the start of every time you run the application. But what it's going to do is check in the left-hand panel for this assistance.json file. If there is none there, it will try to create a new assistant. Every assistant that you create on the OpenAI servers costs you an API usage. So you don't want to have creating a new assistant every time you run the app. So what this does is if there is no assistant file available, it's going to go through the creation process and look at, okay, what model am I using? Okay, what tools am I using? Okay, I have this knowledge base. I have these two functions, et cetera. And it will sort of combine all those ingredients and then... push that up into the servers create the assistant and then what it sends back is the assistant id and that assistant id is saved to this json file and at the end of this function it is sent back to the main function if there's already assistant created it's just going to load it from the file which is great we're not using unnecessary amounts of api usage but if there isn't it's going to create a new one and save that for next time what this means for you is that if you want to recreate your assistant and you want to change the configuration say you want to change the knowledge base or you want to change how the functions operate what you need to do is delete this assistant.json file and then run the app again and it's going to create it with the new settings so just so you understand how i'm managing the assistant id and creation of assistants in this application if you did everything right you should have this down in the bottom left corner where it says it's running now and this window would have popped up in the web view but it says not found it's not an error don't worry this is how it's supposed to be but what you want to do is click on this new tab button click there and it's going to pop up with this url here so this is where our web server is being hosted this is how we can interact with the application we've just built but we're not going to do it through a browser we're going to do it programmatically within voiceflow which brings us to our next step in this build which is to sign up to voiceflow if you haven't already there's going to be a link in the description if you want to sign up there this is my favorite bot building platform and this is what i build a lot of the stuff i build at my development company if we're going to go sort of the lower code route this is the best platform that we have found for creating these kind of builds i really love using this platform and highly recommend it if you're going to be doing anything like this once you've logged in you should see a screen like this Now what you're going to need to do is go to my resource hub and download the voice flow template So i've already created a template for you That does everything you need to do out of the box as you're going to see in a second But in order to get that you need to sign up to my resource hub It's free and you can download that file Then once you've downloaded the file You need to come back to voice flow and click on this import.vf file And in your downloads, you should be able to select your voice flow file In this case, there's one down here when you click that and click open It's going to pop up in the corner and ask if you want to open the assistant which you reply yes to and then you will see a screen that looks a little bit like this so this is as simple as It is for us to be able to deploy all of our functionality that we've created and all of the Assistant API magic to a website and to a website widget chatbot. So just to walk you through the Voiceflow build quickly, we have firstly here our create thread. So this is going to be where we paste in our URL. If you come back to this page here, come and copy this. And then to replace this, you just need to go up to the start. You should have slash start at the end of your URL. Make sure it's exactly like that. It's your URL slash start. That's all you need. and then you can come to this one as well generate response and do the same thing replace this url with your one and then we have everything set up walk you through things again this is where we create a thread by calling this url we're going back to our replit here which has this function in our main.pi which is the slash start so this is going to create a new conversation this is going to create a new thread in our assistant and then we're going to be able to say hey i'm lucy smith solar assistant i can help you with any questions about solar for your home I can also calculate your potential solar savings. How can I help you today? Then we're going to wait for the user's input. We're going to capture that and we're going to send that off to our other endpoint which is the slash chat and then we have it here which pulls out the message from the user, passes it to our assistant and it does all the magic and here you can try to dig through it. It's quite complex and for those of you who ask me why I use Replit for these kind of builds is because look at all of this. I would much rather have this on a Replit where I have full control over and I've really understand the environment rather than trying to cram all of this into voice flow and do that multiple times at different stages in the app so i keep it all here because we can sort of tie our ends make it look all pretty and then just access it through one simple um api endpoint and call this within voice flow so that's the chat functionality what we do is send off the thread id that we created earlier so when we do this we're going to save the thread id that we get back from the api and then we send it off with the last utterance which is what the user said So we send off the user's message, we send off the thread ID to our endpoint and it processes it and sends the AI response back, which we save as the last response. And then we come down here. This is a little snippet that someone from my comment section gave to me. Instead of using a AI task to remove this, we can actually just remove it using regular expressions, which is great. Huge actually got it working. So this is saving us on AI tokens, which is an important thing that many of you brought up. If you are using something like VoiceFlow, there's a lot of questions about, yes, the $50, but also the token cost if you go over. Because we have removed the AI task from this now, you are not going to be using the AI task down here, which will save you a ton on token costs. Last but not least, we take that trimmed response that's removed any of the citations because when you talk to the knowledge base, sometimes it will send citations and they look kind of ugly, so we want to get rid of them. Then it's going to send that last response as the answer and then it's going to loop back and wait for the next user response and it goes through this loop, kind of up here, down here and back over and over and over again until the end of the conversation. In order for us to test this, we need to make sure that firstly, we have changed our... urls over as i mentioned before so it should just be your replet url taken from here and with the slash start on this one and then on this one it should be slash chat everything else should remain the same go back to your replet make sure it is running and everything is all good here this thing should say stop and there should be a little cursor here then we can click on the run button on the right here and click run test How can I help you today? What are the best solar panels? So I'm going to ask to the knowledge base and there we have it. We're getting responses from the knowledge base. While we're at it, we can test the solar calculator feature. So let's go. I want to calculate my savings. So let's ask it and get it to send back. So this is why they're so cool and so powerful, these particular systems. Prior to this technology coming out and this new release, Getting these kind of interactions that are really natural and flexible between user and the AI was quite difficult to say the least. And look at this, I can ask it about one of the tools it has, then it can come back to me and say, sure, these are the things I need. And then if I just give it partial information, it's going to be smart enough to ask me again for the rest of the information and then pull all of the information it needs out of multiple messages and send it away to our tools, our functions, our custom actions within our backend and then give us the answer. So it's really amazing stuff. Hugere we can go. Could you please provide your address and your average monthly electricity bill? So let's go get a new place here in Florence somewhere. Let's get this guy. Oh, maybe this one won't work, but let's give it a go. Bang, and then it say, I pay $350 per month. Now, if you were watching this and you want to debug and understand what's happening, I've set up print statements here so that you can see what's going on with the application while it's sort of doing the buffering. Now, if we scroll down a bit here, you can see that's the prior response. Hugere you can see it received message. So that's us sending a message to the chatbot, calculating solar potential. Oh, where's it gone? Calculating solar potential for that address with bill amount three. 50 so it's taken its passed through the message and has pulled out the address and the bill amount so it's separate variables Then it's converted that to coordinates so you can see and then it's retrieved the soldered data for that location successfully Then we've simplified the data that we get back from the API and then we have the assistance response here, which is really handy So that's printed out the exact response that gets sent in here So if you come back you can see that we have the proper response here So seems like this guy has a $20,000 federal incentive available Total cost without solar will be $100,000 over 20 years. The API gives us back information about the cash purchase option, finance purchase option, and leasing option, and then gives a bit of a summary at the end, thanks to our prompted assistant, knowing what to do with the information. And if you look carefully at the end here, you can see that it is asking me for my information, my name and phone number, so that one of the team members can get in touch with you. This is the lead capture step. So all I need to do is go, let's call myself Huge Jackman. Phone number So it's going to take that name. Oh, Huge Jackman, not Huge Jackman. Forgive it a second. Yep, thank you, Huge. One of our team members will reach out to you shortly at the number you provided. Now we should see in our Airtable, Huge Jackman. Boom, Huge Jackman. Phone number and it's also pulled their address so that we have this information. I did also want to add another column. That was information, so it could pull some of the savings data from the output that we got from the solar calculator and use this to sort of fill out the lead information. It could also take some of the information from the conversation. Say they asked a question about, hey, what's the best panels? What's the best financing option? We could get the AI to analyze the conversation, the thread, and pluck out the kind of questions or concerns that they might have so that we have a little bit more data if our guys are going to go out and reach out to these people. So, okay, we have the address, we have the questions they were asking. we have the phone number we have their name a ton of good lead data and we also have a little status column here so so if i reach out to them we can be interested not interested etc and then we have a working demo of our assistant within voiceflow the last step now is to export this functionality to a website via a little embeddable script we can add to any website which i'll show you in a second but if you are enjoying the video so far i'd really appreciate if you give me a like on this i put a ton of work into this to bring the latest and most exciting things that i can find that i'm working on right now and giving them away to you for free in the form of a tutorial and template as well so if you are enjoying want to support the channel please leave a like right now it would mean a lot to myself and the community alrighty now the last step we need to do is click on this integrations tab on the left click here and you should see these three boxes pop up on the right here is an example of how it's going to look on a website so at the moment i'm happy with the styling i believe the styling should carry over when i export this template we have the installation script which you're going to grab in a second you have the general so i've got the name description i've got the chat persistence set to forget after tab is closed and reloaded just so that i can run through it a bunch of times you'd probably want to see it to never forget on the setting here but for the purposes of testing i'm going to leave it there then we have the appearance section so i've modified this a bit with the color logo etc i'm pretty happy with that so we can click copy here and in this case i've got a demo website as i did in my recent video i've just pulled a free template off the internet somewhere that looks kind of similar to a solar company and all you need to do to add it into any website is to come down to the bottom and add it as a script tag. So here I'll delete the old one. At the bottom of most websites, you should see a stack of these script tags. All you need to do is sort of jump in between a couple of them just to be safe and then paste in that script tag there. We can save that. I can run the website. If I refresh it, there you can see I have the webpage and at the bottom right, I have our chatbot. And there we have it. Hi, I'm Lucy Smith, Solar Assistant. I can help you with that. We can test it again. What are my financing? Options and then we have it with our cash purchases, loans and financing, solar leases and power purchase agreements. Great we can go my address is and then we can pull in this guys up here. 5333 Schultz Street is and I pay $400 per month. As expected it's given us back the information we requested. Federal incentive, cost without solar, some of the leasing, cash purchase and finance options etc. It's asking for my name and phone number so I can go Bruce Willis and my phone number is pop back over to Airtable and there we go Bruce Willis with his address and his phone number has all been logged correctly into our Airtable CRM and there is one final thing that I want to mention about this kind of build which is by running it on Replit like this and just running it within the editor You're only running a very basic and limited server and infrastructure to manage this application. So what you can do within Replit, they actually have a deployment option. So you can go down in the bottom left here and click on deployments. And there's going to be deployment options on the side now. You can pick whichever one you like to do. Go through the setup, so set up your deployment. You can pick how much money you want to spend. Let's say we just go one machine, we can go through. Essentially, you can go through the steps here and fill this out and click on this deploy button. Then it's going to actually deploy your application to the servers that can run constantly that are going to allow you to call this whenever. So when you're taking it to production grade, there is another step required. But in this video, I want you guys to just be able to spin it up as quickly as possible. If you want me to go in depth on deployments, and how you can manage a production grade application rather than just a demo or a prototype like this, then let me know in the comments below and I can definitely get that out to you. That's it for the build. I hope you've been able to follow along. I hope you see how insanely powerful these things are as a lead generation tool in this case, also helping customers learn more about the business, learn more about the products they're selling and being able to flexibly use these different functions and pull in information from the chat. It's just so much better than the kind of stuff we've had to work with before. Prior to this assistance release, getting some of this functionality was incredibly important. close to impossible now we can do it in a couple of hours if you do enjoy these kinds of videos and want an a to z course on how you can create these kind of chatbots and assistance for your business or to sell to other businesses be sure to subscribe to the channel because i have a new video coming breaking down things step by step i know i've been kind of fast on some parts of this but really the entire process for creating these functions step by step and we're going to be using a ton of examples in this video so if you want to be in the loop and not miss that video be sure to subscribe to the channel down below Or if you can't wait for that, you can join my AI Business Accelerator where we have some of the sharpest AI entrepreneurs in the space right now, all moving on this opportunity. And we're all sharing knowledge together, sharing learnings. Plus, I'm in there doing full workshops on this kind of stuff multiple times a week. So if you want to get in there, there'll be a link in the description for you to join the community. Of course, as mentioned, all templates and resources from this video are available in the description by my free resource hub. So you can head down there and check those out. The Replit template, the Voiceflow template, the Airtable template. and the knowledge base document are all going to be there for you guys to take and copy this exact build in just a few minutes i also have my free telegram where i'm dropping daily information and learnings on the gpt's topic so if you're not already in the telegram you can join that in the description as well if you want to learn more about how you can make money by selling these kind of gpt solutions to businesses you can check out my video here where i break down the best ways to make money with gpts but aside from that that's all for the video guys thank you so much for watching and i will see you in the next one

So by the end of this video, you'll not only know how to build your own custom GPTs using the assistance API and put them on your websites, but you'll also know how to plan, create and implement insanely powerful custom actions into your GPTs to create real value for your businesses or for your clients. And at the end of the video, I'll be giving away everything that I'm building today as a template so that you can steal this insanely powerful lead generation chatbot and start using it in your own business or selling it yourself. If you're new to the channel and don't know who I am, my name is Liam Otley and I run my own six-figure development company called Morningside AI. And I also run the largest community of AI agency owners in the world where my students learn how to make money by selling AI solutions to businesses Just like the lead generation chatbot that you're about to build.

Let's get into it. To kick things off, I'm actually going to be showing you the finished product here on screen so that you get an idea of where we're going with this and how much value we are creating when we do this right. So on screen is a template website that I've downloaded.

Just refreshing the page so you can see that it's live. If I click on the widget in the right here, up pops the chat window and I can start talking to it. Hi, I'm Lucy, Smith's solar assistant.

I can help you with any questions on solar for your home. I can also calculate the potential solar savings for your property. How can I help you today? That gives you an idea of the two functionalities of this chatbot.

So firstly, we can ask questions to the knowledge base. What is the cheapest type of panel? This will be querying our knowledge base document, which you guys will get to see in a second.

The general FAQs and information brochure for anything related to solar for single family homes. Hugere we have the answer. The cheapest type of solar panel is polycrystalline solar panel. Great. So that shows the knowledge base is working, but more importantly, here's the feature.

Hugelp me calculate my solar. Solar savings. Hugelp me calculate my solar savings please.

The assistant is going to pick up on this and going to realize that we need two inputs which is going to be their home address and their monthly electricity bill in USD. So for this I'm picking someone's lucky address in the United States. So let's grab this one here, click on this, sorry to whoever's home this is, broadcasting your address out, copy this, head back and if I just put in just the address here it should ask me for the monthly bill as well because it needs the two inputs to calculate this.

Great. Thanks for sharing your address. To proceed, I'll also need your average monthly electricity bill. So let's say I'm spending $250 per month. Now it's going to take that address, it's going to take that monthly cost of power, and it's going to call the Google Solar API, bring it back, and tell us exactly what kind of savings we can expect for our property if we choose to convert to solar.

There we have it. It's given us a full breakdown of our options for solar based off allocation. I'll break this down a little bit more later.

It's really, really complex stuff that's going on behind the scenes here. But essentially, it's giving us information like... our payback period and the 20-year savings we're going to experience, any federal incentives that are associated with our location. But that's not even all this thing does. At the end, it asks for a lead capture.

So it's asking me for my phone number and my name. And then if I provide that, I go Liam Otley, and I give it a phone number, phone number, and I give that to them. Then what this is going to do is take that information. and it's going to also take my address and it's going to push that all to a crm which i'm going to show you how to set up later and then we have a thank you your details have been passed to our team someone will be in touch with you soon so that's the breakdown and a demo of the bot but before we get into the building i need to do a little bit of explanation so that you're fully clear on the difference between creating gpts as you are on chat gpt and creating your own cool ones within the chat gpt website and the kind of thing that we're building over here which is using the assistance api these are your two different ways of creating gpts now on the chat gpt builder they are locked inside of that With the interface, you can't take what you create on the ChatGPT site when you build these things and then go and slap it on your website or put that on WhatsApp or put it wherever you want. You are locked into that and it's intended for you to just use it in there as a handy little kind of personal tool.

I've said that GPTs in this way are like consumer toys. They're focused for the mass market and consumers to engage and have fun with and build these GPTs. But the other side of this is using the assistance API. I've been over this recently in my other videos, but it's important that you understand the difference.

and that most of the value is going to be unlocked through this assistance api because this allows businesses and anyone really to take that same gpt functionality that you create there which is really cool but then take that and put it wherever you want integrate it into any application any website any voice assistant whatever you want so what i'm teaching you in this video is how to use that side of things the assistance api which is far more powerful and is going to create far more value in our world and really be a much more valuable skill than being able to make a little fun gbts on the chat gbt website then we need to get into a little bit of a deeper breakdown of what we're building and how we're going to build it so if you haven't figured out already we're making a lead generation chatbot for solar companies that's going to be put on their website which is intended to help leads and people who are interacting with the website to get you answers to their questions. Firstly, if they have any questions related to, hey, what solar panels are best? What financing options? What's this?

What's that? Anything that they want to ask as if they were talking to a solar expert, we're going to allow them to ask that to the chatbot, which can respond from a knowledge base using a document that we're provided containing this kind of information. Secondly, the chatbot will also allow them to calculate their potential solar savings if they were to install a system.

And we're going to do that by taking in their address and their monthly power bill and sending that off to Google Solar API. and getting the response back from that. And then finally, once I've got that back from the calculator tool, we're going to ask them for their lead information so we can capture the lead and follow them up.

We can break it down into my three ingredients concept for creating GBTs, which is prompting, knowledge, and actions. The prompting is going to tell it that it is an assistant for Smith Solar, that it can answer questions from a knowledge base regarding solar questions, and it can also provide calculations, and it can also capture the lead at the end of it. So that's all the information that we're stuffing into the prompt.

For the knowledge component, we are uploading a single document containing information, a ton of information on solar for single-family homes and finally the actions are going to be calculate solar savings using the solar api and then capture the lead and send it to airtape the stack of technology that we're using to create this assistant is firstly the assistance api of course the solar api and geocoding api from google replit to host it and voice flow to be the front end for it okay to get started you need to fork my replit template which is going to be available in the first link in the description you can sign up to my resource hub it's 100 free and there'll be a link in the resource hub for this particular video where you can get access to this replica now when you click on the link it's going to take you to this screen and what you're going to want to do is click on the fork button on the right here and we're going to give it a name solar lead gen chat bot and we can fork it if you give it a second it will boot up in here Then you'll be greeted with this here. For those of you who aren't developers or don't know how to code, this might seem a little bit overwhelming, but don't worry, I've set this up in a way where you don't need to do any any major changes to get the kind of results you're looking for. And any ones that you do, I'll point them out very clearly.

And it's just a couple of characters or words that you need to change. So we have our main.py file here, which has our endpoints, the chat endpoint and the start endpoint. Don't worry too much about this.

And we have the functions.py file, which has all the specific functionality, such as creating a lead and air table, such as pulling the data from the Solar API. Etc. So all of the fun stuff happens in here then at the bottom we have our create assistant function I have gone pretty heavy on the comments throughout this file So when it comes to modifying it and changing the knowledge or the functions There should be plenty of comments to help you figure out what you need to do But I'll take you through a crash course now because I know many of you want to take this change the knowledge change the function So this is how you can do it So for changing the knowledge, which is the easiest one to do we have the knowledge dot doc X Which is referencing the document on the left here.

So if you want to change the document you can delete that and you can upload a new one by clicking this. So you can upload a new file and just make sure that the name of the file is referenced here exactly. If you want to add multiple files, you'll need to copy and paste this multiple times and save them to different files.

I recommend that you paste this entire function from the top down into ChatGPT and say, hey, I'd like to add multiple files because it needs to add a couple of different rows here, duplicate that. And then down the bottom here, we need to pass in more than one file ID. So that's for changing the knowledge.

For changing the instructions, we have this instructions assistant instructions variable reference here. And what this is actually doing is pulling from up here, we have an import from our prompts file. And on the left here is the prompts file that we're talking about. So the assistant instructions and the prompts.py. If you want to change the instructions for your assistant, this is where you need to do it.

So by changing this, you can take a look at how to set this thing up here. But by changing this, you're going to change the prompting and the behavior of the chatbot. And the information in here, as you look at my stuff, you'll see that it's referencing how to use the knowledge.

It's referencing how to use the tools and what to do after. So one of the cool things that I added in here was that after the assistant has provided the user with their solar calculations, they should ask for their name and phone number so that one of the team can get in contact with them about installing solar for their home. So that is instructing it to after you've triggered that tool, then do this, which is what you'll see later in the app where it actually asks for the lead capture at the end of the response from the tool.

So if you want to change the prompting and instructions, you need to come to the prompts.py file, modify that. If you want to change the knowledge, you need to come to the top of the create assistance function and modify this and modify your documents. And when it comes to modifying the functions, it's definitely the harder part, but I'll try to walk you through it as quickly as possible now and make it as simple as possible.

So from a high level, we have in the setup for our system, we have our different functions. So again, this bit here is the knowledge tools. So we see retrieval means that it's adding the knowledge base retrieval tool, which is going to handle the knowledge base up here.

But the one below it is a function. So we have our solar panel calculations and then we have the properties and the inputs it provides. So we need the address and we need the monthly bill. And both of them are required. And the same down here.

We have another function called create lead, which is going to log the lead in the database. It has a little description and then it's going to need the name, the phone, the address, and all of them are required. Now, you don't need to write these all yourself.

You can ask ChatGPT to assist you with this and by pasting in this entire function, it's going to show them exactly how you need to do it. So this is your reference and ChatGPT can help you to modify it. Now when it comes to actually what functions are being used, the name of the function here needs to match the name of the function that you want to call. In this case, I want to call this, which is my sort of full calculations function.

Again, if I go down to here, the name of the function is called create underscore lead. And at the top, I have create underscore lead. Now this all comes together. when you go to the main.py file.

So over here, if I scroll down to the chat endpoint and I scroll down to this section, you'll see in this part here, if the assistant picks up action required, meaning one of the tools has been triggered based off the natural language description of capture the lead details, calculate the solar potential. If one of those are triggered, it's going to come down here and look up for the different function names. So we have the solar panel calculations and the create lead.

When the assistant initiates the action, it's going to pass the name through. and we can go okay what's the tool call that it's called what's the name of that okay it matches the solar panel one great then we want to call from our functions file the solar panel calculations and that takes the address argument and it takes a monthly bill argument you can see this back here it takes an address and a monthly bill and in this case we are just pulling from the tool call we are pulling the address as the first parameter and the monthly bill is the second parameter so Might seem a bit complicated, but again, if you paste this in the chat GPT, it should understand what you're trying to do. Same thing happens down here with the create lead. Again, we needed to have three parameters in here, which is the name, phone, and address. And if the function name that was matched within the tool matches the create lead name here, then it's going to trigger that and pass the arguments in and then output it.

That might be a little bit complex for some of you, but I did promise that you would understand how to put these things together and how it all fits. You might need to go back and break that down a couple of times, but that is how it works. The next step for you to set this up is to head over to Google Cloud. So cloud.google.com and you need to come on here and create account. I assume you've already got a Gmail account and you can click on go to console.

You might need to do a bit of set up and sort of creating an account on the Google Cloud platform. Once you're in, you should see this. And then what we need to do is come up to the top left. Let's create a new project. I'll call this solar.

chat bot for this i'll be using my youtube projects billing account in your case you'll need to set one up just so you're clear on the amount of cost i've been using this a decent amount and i've been charged under a dollar so these things are very cheap to use for the api usage so create your billing account you need to put a card on it so that they're able to verify it and allow you to start using it but once you've set up your billing account and put a card on file you'll be able to select it here and then for the organization i'll just leave it as none so it's going to take a second for it to create my project then i can click select project here so that i jump into it then once it's all set up what you can do is click on the apis and services section here and go to library and then what you want to do is search up geo coding geocoding click on the geocoding api need to enable that when you successfully enable that it should pop up with this api key message so you can copy your api key here which is what we're going to need to use in our application to be able to call the solar api So we can copy this, save it somewhere and go continue to Google Maps. Maybe later for the protecting my API key. Then we'll go back to APIs and services. Let's close this on the side.

Then you'll see the solar API appear here. So we can click enable on the solar API. Give that a second.

and we can scroll down to make sure we have the geocoding api enabled great we have our google cloud set up we have our api key and we have the solar api and the geocoding api enabled you want to grand bit api key and come back over to our replit and if i scroll to the secret section and click on secrets you should see the google cloud api key section here what you want to do is click the edit button here and then paste in your api key here and click update secret next we need to set up our air table to log all of the leads which will act as our temporary crm and we're you can use obviously any kind of crm that you like i just like using air table because i'm very familiar with the api so in this example we'll be using air table i do have a base here that you'll be able to copy if you want to get access to this crm you can copy and clone it over to your account it'll be in the resource hub as well so there'll be a link to click on that then you should see a page like this and you will see this button here that says copy base so if you copy that as your air table account you'll have this all set up as i've got it done here then we need to do one more thing which is to go on to our developer hub on the right here and you will see all of our personal access tokens so you'll need to click on create token up here on the right give it a name we'll call it lead gen chat bot not the best name and we can go to scopes and we need to add data records read data records right and then we need to set up the access so in this case i'm going to go down to smith solar crm and add that and i can create the token it'll then pop up with your api key which you need to copy and then bring back over to replit again and do the same process so click here edit and then paste it in and save the updated API key. The final step for setting up these keys is to get your OpenAI API key. To create or copy your API key for OpenAI you can come to platform.openai.com slash api dash keys and on the page here you will see a button to create a new secret key.

You can click this button it'll give you a new key you can copy that back over to here and do the same process of updating the API key. Now I will mention I had a bunch of people asking in the comments that they couldn't get access to the model that we are using in this particular build. That's because you likely need to put your card on file.

What you need to do is go to the settings on the side here, you need to add a card for billing. You're not necessarily going to be billed for it because you do have that free allowance but you need to put a card on file so that they allow you to use these models. Just a little issue I saw in the comments of the last video. The very last bit of setup we need to do here is to get our Airtable base URL.

So what you can do is click on the profile icon in the corner, click on developer hub. and then we can click on web api documentation and then you can scroll down and click on your smith solar crm and that will take you to the documentation for your particular base which is super handy if you scroll down to the leads table in this case we want to go to Create leads record and it will give you a URL here So we want to copy this all the way up to the end of the leads word and we've already set this up when I Wrote the code this is already kind of integrated into the function All you need to do is head back and come to this create lead function and replace that entire Bit there I already have mine in there so I didn't change anything But you should replace this URL with your URL with all of that done. We are pretty much ready to boot this thing up Let's just save this file. Make sure we've all saved up. Let's go back and give the main.par file a save as well And we can click this big juicy green run button on the top.

Give it a second and it's going to boot up. On the left panel here, you'll see that there is no assistance.json and you'll see a file pop up here in a second. So give it a sec.

And then we have the assistance.json file pop up. The way that I've set this up is that the create assistance function will run at the start of every time you run the application. But what it's going to do is check in the left-hand panel for this assistance.json file. If there is none there, it will try to create a new assistant. Every assistant that you create on the OpenAI servers costs you an API usage.

So you don't want to have creating a new assistant every time you run the app. So what this does is if there is no assistant file available, it's going to go through the creation process and look at, okay, what model am I using? Okay, what tools am I using?

Okay, I have this knowledge base. I have these two functions, et cetera. And it will sort of combine all those ingredients and then...

push that up into the servers create the assistant and then what it sends back is the assistant id and that assistant id is saved to this json file and at the end of this function it is sent back to the main function if there's already assistant created it's just going to load it from the file which is great we're not using unnecessary amounts of api usage but if there isn't it's going to create a new one and save that for next time what this means for you is that if you want to recreate your assistant and you want to change the configuration say you want to change the knowledge base or you want to change how the functions operate what you need to do is delete this assistant.json file and then run the app again and it's going to create it with the new settings so just so you understand how i'm managing the assistant id and creation of assistants in this application if you did everything right you should have this down in the bottom left corner where it says it's running now and this window would have popped up in the web view but it says not found it's not an error don't worry this is how it's supposed to be but what you want to do is click on this new tab button click there and it's going to pop up with this url here so this is where our web server is being hosted this is how we can interact with the application we've just built but we're not going to do it through a browser we're going to do it programmatically within voiceflow which brings us to our next step in this build which is to sign up to voiceflow if you haven't already there's going to be a link in the description if you want to sign up there this is my favorite bot building platform and this is what i build a lot of the stuff i build at my development company if we're going to go sort of the lower code route this is the best platform that we have found for creating these kind of builds i really love using this platform and highly recommend it if you're going to be doing anything like this once you've logged in you should see a screen like this Now what you're going to need to do is go to my resource hub and download the voice flow template So i've already created a template for you That does everything you need to do out of the box as you're going to see in a second But in order to get that you need to sign up to my resource hub It's free and you can download that file Then once you've downloaded the file You need to come back to voice flow and click on this import.vf file And in your downloads, you should be able to select your voice flow file In this case, there's one down here when you click that and click open It's going to pop up in the corner and ask if you want to open the assistant which you reply yes to and then you will see a screen that looks a little bit like this so this is as simple as It is for us to be able to deploy all of our functionality that we've created and all of the Assistant API magic to a website and to a website widget chatbot. So just to walk you through the Voiceflow build quickly, we have firstly here our create thread. So this is going to be where we paste in our URL.

If you come back to this page here, come and copy this. And then to replace this, you just need to go up to the start. You should have slash start at the end of your URL.

Make sure it's exactly like that. It's your URL slash start. That's all you need. and then you can come to this one as well generate response and do the same thing replace this url with your one and then we have everything set up walk you through things again this is where we create a thread by calling this url we're going back to our replit here which has this function in our main.pi which is the slash start so this is going to create a new conversation this is going to create a new thread in our assistant and then we're going to be able to say hey i'm lucy smith solar assistant i can help you with any questions about solar for your home I can also calculate your potential solar savings. How can I help you today?

Then we're going to wait for the user's input. We're going to capture that and we're going to send that off to our other endpoint which is the slash chat and then we have it here which pulls out the message from the user, passes it to our assistant and it does all the magic and here you can try to dig through it. It's quite complex and for those of you who ask me why I use Replit for these kind of builds is because look at all of this.

I would much rather have this on a Replit where I have full control over and I've really understand the environment rather than trying to cram all of this into voice flow and do that multiple times at different stages in the app so i keep it all here because we can sort of tie our ends make it look all pretty and then just access it through one simple um api endpoint and call this within voice flow so that's the chat functionality what we do is send off the thread id that we created earlier so when we do this we're going to save the thread id that we get back from the api and then we send it off with the last utterance which is what the user said So we send off the user's message, we send off the thread ID to our endpoint and it processes it and sends the AI response back, which we save as the last response. And then we come down here. This is a little snippet that someone from my comment section gave to me. Instead of using a AI task to remove this, we can actually just remove it using regular expressions, which is great. Huge actually got it working.

So this is saving us on AI tokens, which is an important thing that many of you brought up. If you are using something like VoiceFlow, there's a lot of questions about, yes, the $50, but also the token cost if you go over. Because we have removed the AI task from this now, you are not going to be using the AI task down here, which will save you a ton on token costs.

Last but not least, we take that trimmed response that's removed any of the citations because when you talk to the knowledge base, sometimes it will send citations and they look kind of ugly, so we want to get rid of them. Then it's going to send that last response as the answer and then it's going to loop back and wait for the next user response and it goes through this loop, kind of up here, down here and back over and over and over again until the end of the conversation. In order for us to test this, we need to make sure that firstly, we have changed our... urls over as i mentioned before so it should just be your replet url taken from here and with the slash start on this one and then on this one it should be slash chat everything else should remain the same go back to your replet make sure it is running and everything is all good here this thing should say stop and there should be a little cursor here then we can click on the run button on the right here and click run test How can I help you today?

What are the best solar panels? So I'm going to ask to the knowledge base and there we have it. We're getting responses from the knowledge base. While we're at it, we can test the solar calculator feature.

So let's go. I want to calculate my savings. So let's ask it and get it to send back.

So this is why they're so cool and so powerful, these particular systems. Prior to this technology coming out and this new release, Getting these kind of interactions that are really natural and flexible between user and the AI was quite difficult to say the least. And look at this, I can ask it about one of the tools it has, then it can come back to me and say, sure, these are the things I need. And then if I just give it partial information, it's going to be smart enough to ask me again for the rest of the information and then pull all of the information it needs out of multiple messages and send it away to our tools, our functions, our custom actions within our backend and then give us the answer. So it's really amazing stuff.

Hugere we can go. Could you please provide your address and your average monthly electricity bill? So let's go get a new place here in Florence somewhere.

Let's get this guy. Oh, maybe this one won't work, but let's give it a go. Bang, and then it say, I pay $350 per month. Now, if you were watching this and you want to debug and understand what's happening, I've set up print statements here so that you can see what's going on with the application while it's sort of doing the buffering. Now, if we scroll down a bit here, you can see that's the prior response.

Hugere you can see it received message. So that's us sending a message to the chatbot, calculating solar potential. Oh, where's it gone? Calculating solar potential for that address with bill amount three.

50 so it's taken its passed through the message and has pulled out the address and the bill amount so it's separate variables Then it's converted that to coordinates so you can see and then it's retrieved the soldered data for that location successfully Then we've simplified the data that we get back from the API and then we have the assistance response here, which is really handy So that's printed out the exact response that gets sent in here So if you come back you can see that we have the proper response here So seems like this guy has a $20,000 federal incentive available Total cost without solar will be $100,000 over 20 years. The API gives us back information about the cash purchase option, finance purchase option, and leasing option, and then gives a bit of a summary at the end, thanks to our prompted assistant, knowing what to do with the information. And if you look carefully at the end here, you can see that it is asking me for my information, my name and phone number, so that one of the team members can get in touch with you.

This is the lead capture step. So all I need to do is go, let's call myself Huge Jackman. Phone number So it's going to take that name.

Oh, Huge Jackman, not Huge Jackman. Forgive it a second. Yep, thank you, Huge. One of our team members will reach out to you shortly at the number you provided.

Now we should see in our Airtable, Huge Jackman. Boom, Huge Jackman. Phone number and it's also pulled their address so that we have this information. I did also want to add another column.

That was information, so it could pull some of the savings data from the output that we got from the solar calculator and use this to sort of fill out the lead information. It could also take some of the information from the conversation. Say they asked a question about, hey, what's the best panels?

What's the best financing option? We could get the AI to analyze the conversation, the thread, and pluck out the kind of questions or concerns that they might have so that we have a little bit more data if our guys are going to go out and reach out to these people. So, okay, we have the address, we have the questions they were asking.

we have the phone number we have their name a ton of good lead data and we also have a little status column here so so if i reach out to them we can be interested not interested etc and then we have a working demo of our assistant within voiceflow the last step now is to export this functionality to a website via a little embeddable script we can add to any website which i'll show you in a second but if you are enjoying the video so far i'd really appreciate if you give me a like on this i put a ton of work into this to bring the latest and most exciting things that i can find that i'm working on right now and giving them away to you for free in the form of a tutorial and template as well so if you are enjoying want to support the channel please leave a like right now it would mean a lot to myself and the community alrighty now the last step we need to do is click on this integrations tab on the left click here and you should see these three boxes pop up on the right here is an example of how it's going to look on a website so at the moment i'm happy with the styling i believe the styling should carry over when i export this template we have the installation script which you're going to grab in a second you have the general so i've got the name description i've got the chat persistence set to forget after tab is closed and reloaded just so that i can run through it a bunch of times you'd probably want to see it to never forget on the setting here but for the purposes of testing i'm going to leave it there then we have the appearance section so i've modified this a bit with the color logo etc i'm pretty happy with that so we can click copy here and in this case i've got a demo website as i did in my recent video i've just pulled a free template off the internet somewhere that looks kind of similar to a solar company and all you need to do to add it into any website is to come down to the bottom and add it as a script tag. So here I'll delete the old one. At the bottom of most websites, you should see a stack of these script tags. All you need to do is sort of jump in between a couple of them just to be safe and then paste in that script tag there. We can save that.

I can run the website. If I refresh it, there you can see I have the webpage and at the bottom right, I have our chatbot. And there we have it. Hi, I'm Lucy Smith, Solar Assistant. I can help you with that.

We can test it again. What are my financing? Options and then we have it with our cash purchases, loans and financing, solar leases and power purchase agreements.

Great we can go my address is and then we can pull in this guys up here. 5333 Schultz Street is and I pay $400 per month. As expected it's given us back the information we requested.

Federal incentive, cost without solar, some of the leasing, cash purchase and finance options etc. It's asking for my name and phone number so I can go Bruce Willis and my phone number is pop back over to Airtable and there we go Bruce Willis with his address and his phone number has all been logged correctly into our Airtable CRM and there is one final thing that I want to mention about this kind of build which is by running it on Replit like this and just running it within the editor You're only running a very basic and limited server and infrastructure to manage this application. So what you can do within Replit, they actually have a deployment option.

So you can go down in the bottom left here and click on deployments. And there's going to be deployment options on the side now. You can pick whichever one you like to do.

Go through the setup, so set up your deployment. You can pick how much money you want to spend. Let's say we just go one machine, we can go through. Essentially, you can go through the steps here and fill this out and click on this deploy button. Then it's going to actually deploy your application to the servers that can run constantly that are going to allow you to call this whenever.

So when you're taking it to production grade, there is another step required. But in this video, I want you guys to just be able to spin it up as quickly as possible. If you want me to go in depth on deployments, and how you can manage a production grade application rather than just a demo or a prototype like this, then let me know in the comments below and I can definitely get that out to you. That's it for the build. I hope you've been able to follow along.

I hope you see how insanely powerful these things are as a lead generation tool in this case, also helping customers learn more about the business, learn more about the products they're selling and being able to flexibly use these different functions and pull in information from the chat. It's just so much better than the kind of stuff we've had to work with before. Prior to this assistance release, getting some of this functionality was incredibly important. close to impossible now we can do it in a couple of hours if you do enjoy these kinds of videos and want an a to z course on how you can create these kind of chatbots and assistance for your business or to sell to other businesses be sure to subscribe to the channel because i have a new video coming breaking down things step by step i know i've been kind of fast on some parts of this but really the entire process for creating these functions step by step and we're going to be using a ton of examples in this video so if you want to be in the loop and not miss that video be sure to subscribe to the channel down below Or if you can't wait for that, you can join my AI Business Accelerator where we have some of the sharpest AI entrepreneurs in the space right now, all moving on this opportunity.

And we're all sharing knowledge together, sharing learnings. Plus, I'm in there doing full workshops on this kind of stuff multiple times a week. So if you want to get in there, there'll be a link in the description for you to join the community. Of course, as mentioned, all templates and resources from this video are available in the description by my free resource hub. So you can head down there and check those out.

The Replit template, the Voiceflow template, the Airtable template. and the knowledge base document are all going to be there for you guys to take and copy this exact build in just a few minutes i also have my free telegram where i'm dropping daily information and learnings on the gpt's topic so if you're not already in the telegram you can join that in the description as well if you want to learn more about how you can make money by selling these kind of gpt solutions to businesses you can check out my video here where i break down the best ways to make money with gpts but aside from that that's all for the video guys thank you so much for watching and i will see you in the next one

Transcript for:Building a Custom GPT Lead Generation Chatbot

Transcript for:
Building a Custom GPT Lead Generation Chatbot