How to create an AI chatbot for instagram: Ultimate Guide 2024

One of the things that can really bog you down as a small business owner or someone working for one is messages. I know you may be flooded with tons of Facebook and Instagram DMs, and sometimes it just feels really impossible to be able to keep up. Trust me, I've been there. As soon as you think you got to inbox zero and there's more messages rolling in, and while you're trying to focus on growing your business, it may be just really hard to keep up with the influx of customer requests that you may also be receiving. So in today's tutorial, I'm going to show you how you can make an an AI chatbot that can offer your customers support on all of your social media channels 24/7, 365. This is going to free up so much of your time and offer better customer services to your customers. Also, if you're new to the channel, I would really appreciate it if you can like and subscribe. Also, if you don't have a small business and you want to learn how you can profit from the AI age that we're in, learn this skill set and you can offer to build these AI chatbots for businesses and make thousands of dollars every single month. So without further ado, let's jump in and see how we can do this. So I created a fictional bakery here called Helena's Bakeshop, and I used ChatGPT to write all of the FAQs for this fictional bakery. So you can see here, uh, what is your opening hours? We're open from 7 AM to 7 PM. So as great as ChatGPT is, and it has been trained on, I think right now, over 175 million pieces of information like blog posts, YouTube transcripts, you name it, all over the web, but it doesn't know the specific things about your business. Like your own company policies, refund policies, products that you offer, maybe opening hours, etc. So a request that I see very often from business owners is, how can I leverage this AI technology into my business? How can I create a personal AI assistant that knows everything about my business? And this is where the power of AI can be leveraged for your business. I'm going to show you how to make these AI assistants today. So You can see here this document goes into all of the FAQs about my fictional bakery, from questions like, is there gluten-free products at your bakery? Do you offer loyalty programs? Do you offer gift cards? Do you offer, uh, do you offer parking? What is your refund policy? And so forth and so forth. And if you want a copy of this FAQ so that you can follow along with this tutorial, the link is going to be in the descriptions below. After you have your list of FAQs, we're now ready to go on to our our second step, which is to create our AI assistant. And we're going to do this with the help of OpenAI Assistant. You can get over here by just typing in OpenAI Assistant into Google. It's going to take you to this link, or you can simply go on to platform.openai.com/assistants. It's going to take you to this page as well. You do need to be on the paid version of ChatGPT, which is $20 a month. And if you think about it, that's like 1 hour of what you need to pay a human assistant, and for that price you can get an AI assistant 24/7, 365. So it's really not a big price to pay. And then once you're on the assistant page, you can see all the assistants you created before, but we want to go ahead and create a new assistant right now. I'm going to press the create button on the top left-hand corner— right-hand corner, and Now this is where I can give the assistant a name. I can call it Helena Bake Shop Assistant. And then under, under the instructions portion, this is where I put in the system prompt. This is how I want the assistant to behave, and I describe the role that I want this assistant to take on. I have now put in my instructions for this AI assistant. I'll read it out to you.

I'm here to provide quick and accurate answers to customers' frequently asked questions. I'll assist users in finding About products, services, hours of operation, and other bakery-related inquiries. My responses will be concise, informative, tailored to the bakery's context. I'll ask clarification when the query is ambiguous and aim to ensure customer satisfaction with clear and friendly answers. Avoid giving personal opinions or information outside of my training on Helena's Bake Shop.

This is important here because a lot of business owners come to me and they say, Helena, I don't want the AI to be able to answer any question. If someone asks the AI what's the color of the sky, I don't want it to keep talking to this customer. I wanted to be focused on everything in my business, and that's where creating these custom AI assistants really come and shine. After you have put in your system instructions here, what you want to do is go down to Retrieval and turn this button on. This will allow you to upload specific documentation that you want this AI assistant to know. And, and then under Files, I'm going to press Add And then I'm just simply going to add in the, the text file with all of the FAQs that we have just created. All right, one last thing we need to change in our system to get this working, and that is changing the model. It's very, very important here. We have to choose GPT-3.5 Turbo 0125. The other models are not going to work. So please make sure you switch this over to get this chatbot working. We have now created a custom AI assistant that knows all of the FAQs related to our fictional bakery business. Now we're ready to go on to the next step, which is connecting this AI assistant to the Facebook and Instagram pages that we manage for this fictional bakery. Now the way that we're going to do this is use a tool called ManyChat. If you have ever gone onto Instagram and you've seen those posts where it says, um, reply shop or reply link in the comments and I'll send you the custom link to this product that you have just seen, what you have just seen right there is something called ManyChat at Work. And you can get started with ManyChat for free. The website is manychat.com, M-A-N-Y-chat.com. I'll also include a link to this in the descriptions below. Once you are on your ManyChat account, you're gonna get to this dashboard. And then from here, what you want to do is click on Automations, which is a 3rd tab that you see here. And then from here, we want to go over to the top right-hand corner and click on New Automations. After we have created a new automation. We want to start from scratch, and this is the canvas that we're going to see. No-code tools is all the rage these days, and what it does is instead of you having to code everything, all you need to do is describe what you want to accomplish by dragging and dropping these modules onto the canvas, and you can visually see what's happening. And before we dive into making this automation, let's be clear about what we want to accomplish here. What we want to accomplish is get the user's questions, store that answer somewhere, the question that they want to ask us, and then send it over to the AI assistant that we just created, have the AI assistant come up with the answer, and then, and then display that answer back to the user. All right, now that we know what our overall goal is, let's dive in and let's make this now. So we're going to ignore the trigger for now because we're going to configure it later. So let's go over to the first module or block that we see here, which is sending a message, and we're just going to click in the middle here where it says add a text. And once we click there, we're going to see this left panel show up, and we actually just want to get rid of this text by clicking the X button here. And what we want to do is select user input instead because we want to capture what the user has replied to us with. Okay, that's why we don't want to select text, because if we select text, we're not going to be able to capture their response. Under enter a question, here. This is the, the question that we're asking the user to get what they want from us, right? So we can say something like, hello, how can I help you? Feel free to ask me anything.

Great.

And then from here, we are now capturing what the user says back to us. So when they say something back to us, we want to get it to go over to our AI assistant. So what we're going to do is drag the next step and then drop it anywhere on the canvas. And we're going to see this panel show up, which is basically ManyChat asking us what is going to happen next, right? And we're going to select Action and then press +Action here on the left panel navigation. And then we're going to select Make External Request, okay? Because we have to go externally to our AI assistant to ask that question. After we have selected Make the External Request, we're going to to now configure our Make.com account. Make.com is an RPA or robotic processing automation platform. Again, it's no-code, and what it does is it acts like the glue that glues different applications together. In, in this case, what we wanted to do is connect our ManyChat via a webhook to the AI assistant that we have just created. And you can get started with Make.com for free as well. The link is going to be in the descriptions below. Once you're on your, uh, on your make.com and you have signed up for a free account, you want to go over and select scenarios. After you select the scenarios, you're going to go ahead at the top and create a new scenario here, and then you'll be taken to a blank canvas that looks like this. Again, you just have to describe and think about what is the step of this automation that you wanted to happen and be able to visually drag these modules appropriately onto the canvas to achieve the kind of automation that you want to achieve. And I really think this is such a game changer in the society today because you can automate so many different things which will save you so much time. So we're going to go ahead and press the big plus icon in the center, and it's going to ask us what kind of automation we want to create here. Under the search application, we're going to type in webhook. And what a webhook does is that it listens for activity. And the reason we want to— I want this trigger to be a webhook is because we want it to listen right to ManyChat and see if there's any user input. Is there any user asking us a question? And this automation will only run when that has been triggered, when there has been a user asking us a question. And then once we get to the webhook, we're gonna add a new one and we can give it a name. I'll just call it Bakery Bakery Test, right? And we can go ahead and press Save right here. And now what we're going to do is copy this webhook address and bring it back into ManyChat. Now that we're back into ManyChat, under External Request, this is where we can add our request. This is where we put in the Make webhook URL that we have just created. Now we had to do a couple more configurations to get this webhook working. We can leave the header. Now we're going to go into the body. So all of these APIs talk to each other with a programming language called JSON, J-S-O-N, but you don't need to know any of that. You can just follow along with this tutorial to get this working. And I don't even know how to code. And if I do need to write specific API requests or JSON code, I just use ChatGPT to do it. I'm sure you can do it too. But this is really simple. You do not need ChatGPT. All we're going to do is put in 2 of these curly brackets. And then in the center here, we're going to put the quote, put in input, and then put another quote. And over here, when you press add a field, and then we're going to select last text input and then put another quote here. So what's going to happen here is this— what we're, what we're going to pass to this webhook is a variable called input. And the value of that input is going to be the last text input, which is the last question or the last inquiry that the user has sent over to us. Now we're going to go ahead and press save here at the bottom. Now we need to have somewhere to save that response and be able to reply and show that response back to the user once the AI has replied with it. And the reason— the way that we're going to do that is go down to settings in your ManyChat account And then go all the way to the bottom where you see Fields. We're going to go ahead and press on Fields, and this is where we can configure custom fields inside of our ManyChat account. I'm going to go ahead and press Add User— New User Field here at the top, which is the big blue button, and then I can give this field a name. Essentially here I'm creating a new variable, and this variable is going to be called AI Response, and it's going to be a text field. I'm going to go ahead and press Create.

Great.

Now I have a custom variable created. I'm going to go back to my automation now, the one that I was just making, and then click back on the external request. Click on the external request I just started making, and then go over to the response mapping. From here, I'm going to put in the JSON field the name of the webhook, right? So it's going to be you always start with a dollar sign, then dot. And because we call it bakery test in Make, that is what the JSON path is going to be called. And this is going to be mapped to the AI response. Okay, so once that whole Make automation that we're going to make in just a second is, is finished, we're going to store the AI response that we get back in it in the variable called AI response. I'm going to go ahead and save it here and then go on to the next step. Again, I'm going to drag from that little circle under next step and it's going to give me an arrow that comes out and I can drop it anywhere on the canvas. After I've done that, I'm going to select Instagram and this time I want to select text and then I'm going to click on the 2 curly brackets right over here and then go down to custom user fields and select AI response. So what we have done here is ask the user, how can I help you, captured their question, sent it over to make.com via webhook, and then once we get that response back from Make, we are displaying their question. And there's just one last thing we're going to do inside of ManyChat before we go over and configure make.com. So I'm going to drag over the next step And then from here, we're going to select Instagram again, and then we're going to put in user input. And here we're going to say, is there anything else that I can help you with?

Great.

And we're going to drag the next step back to the Make URL. So from here, once the user sees a response, They're also going to see, is there anything else I can help you with? They can ask another question and then they can just ask as many questions that they want. Now that we have finished the ManyChat portion, let's go over to make.com and finish the automation from there. Now we just have the, the one webhook over here, which is the bakery test. And then what we wanted to do after we have captured the question from that user is sent it over to the assistant. But before we can go ahead and do that, we actually have to go ahead and edit structure of this webhook, and we can do so by clicking on this module and then press on the Edit button. Then toggle on Show Advanced Settings and then go into the data structure and add a data structure. From here, you want to add an item, and then we just want to put in input because this is what allows this webhook to receive the input that we just put into our ManyChat Okay, and we're going to press Save and Save again, and then press OK. Now we can go on to our next module. And now what we want to do is send the question that we just have into the OpenAI assistant that we just created. So under Search Application, we want to search for OpenAI. And, and then we can see a list of actions that we can take, we want to scroll down until we see Send a Message to an Assistant. This is what we want to press. And if this is your first time connecting to the OpenAI module, you'll have to click on Add, put in your username and password to your OpenAI account, and once you do so, you'll be able to see all of the assistants that you have made inside of your account. So I can see a list of all the assistants I have ever made in the past. I'm going to select the one that we just made, and then under Messages, we— once we click here, we'll be able to see the input variable that we have just captured from the webhook, right? And again, remember the input is the name of the variable that we just named to be able to capture the last response from the user from ManyChat. Okay, this is a question that we want to ask this assistant. And now what we want to do is parse this data and then send it right back over to ManyChat. Okay, so in order to parse this data, we want to create a JSON node, and this is really easy. We want to select JSON and then press Create JSON. And then we can add a data structure. And I can just give it a name again called Bakery Test again, and press Save.

Great.

And under the Bakery Test, once I click here, it's going to ask me what JSON am I trying to create here. And I can see all of the information I have from the previous, the previous modules to be able to choose from. And what I want to select is hover over the content, expand upon that, expand upon text, and select value. So I want to select the output that the assistant just created, the text value of the response that we just created. Okay, and then I'm just going to go ahead and press OK. And then lastly, I now need to send this JSON response back to ManyChat. And, and then so the last module I need to create here is another webhook. And I'm going to select webhook response. And the body here would be the JSON string that we have just created. Okay, now just to explain this again, the reason we need to create this JSON is because that is how APIs talk to each other. If we just send back the text back to ManyChat, it's not going to be able to parse it. So we need to turn that text into JSON code and then use a webhook to send it back to ManyChat. Now we need to set our trigger. And how we're going to do this is first we're going to give this automation a name. Let's say AI chatbot will be the name of this automation. And then we're going to go over to settings. And then from here, we're going to select the channel that we have just connected to. In this case, it's Instagram, and we're going to under default reply select the automation that we have just created, and then it'll be AI chatbot. Pick this automation. All right, and this is it. So I'm just going to go back to ManyChat and then set it to live. Okay, so what I did was I asked a number of different questions from my personal IG account to Helena's Bake Shop Instagram account just to see if it is working. And as you can see, it is working properly. So I asked the bot what is the opening hours. And it's telling me that Helen's Bake Shop is open from 7 AM to 7 PM, which if you go back to the FAQ document that we have just created, this is exactly what I told it to say. I've also asked it if there's gluten-free options. And it's telling me yes, there are gluten-free options. And if we go back to the FAQ documents, this is exactly what we told it to say. And I asked it if there's Wi-Fi in the store. And it says that there is Wi-Fi provided to the customers, which if we go back to the documents, we do say that there is Wi-Fi. So if you're a small business owner or a consultant who wants to work with small business owners, or someone who wants to make money off of the AI revolution, please learn this skill set because making these chatbots can save you one, a lot of time and can help you get clients that you can make thousands of dollars from. Now, thank you so much for joining me for today's tutorial. I hope you got a lot of value out of it. Please like and subscribe below for more informative tutorials just like this.