Speaker 1: In this video, I'm going to give you a step-by-step tutorial on how to create an AI research agent that has like hundreds, if not thousands of use cases using Perplexity AI and Meg.com. Perplexity AI is basically like ChatGPT plus Google at the same time and enables us to look at different URLs, get data and insights from it. Then we'll have an army of GPT assistants that will do a lot of the heavy lifting and turn this into something really incredible. You can build this automation in so many different ways for the purposes of this video and this tutorial. Here's the automation we're going to build. So we're going to feed Perplexity AI a prompt and URL. That URL could be a YouTube video, it could be a Reddit post, it could be an interesting article, just something that we think is interesting and we want to create content on. After that, we're going to transpose that into Google Doc as a script that we can use to make content. Then we're going to send it to three different GPT specialist agents that are going to give us three completely different outputs. One is going to turn that into LinkedIn post, another is going to turn it into a tweet and then finally something for Facebook as well. So from just entering one URL, we're going to get four completely different content outputs and you can use this as a research assistant. But on top of that, what Perplexity is also going to do is find as unique data and insights based on the topic that we're discussing that we just could not get through a typical language model. And not only will this enable us to take, let's say, an interesting article, it will enable us to get deeper, more analytical understanding and get it transposed into a post-ready format for whatever we want. And we'll use this as an example but like I said, there's so many different ways you could use this integration to create something epic. But we're going to start from this from the beginning. So the software we're going to use for this is a no-code automation tool called make.com. You might have heard of it, a similar competitor to it is Zapier but they basically do the same thing. They give you this drag and drop build functionality where you can connect 600, 700 different apps from Gmail, Google Docs, you name it. And this is going to be the no-code automation tool that we're going to use and that's not the call out by the way. There's going to be no coding whatsoever this entire video. It's literally drag and drop and just follow the instructions. So when you come over to make, what I'm going to do is basically you come on to the homepage. So we'll discard all these changes to build it from get. And you have this dashboard, it shows you how much, how many operations that you've had and what the cost of that. So all we're going to do is come to this button here called create new scenario like this. And then there we are, we start from the beginning. So the first thing we want to do is add perplexity, okay? So we search for perplexity here. Right now, the first thing you'll see when you do this is this perplexity community tab because you'll need to come onto a third party website to get the perplexity AI integration. So we come down here, we click this install button at the bottom, we leave make, the make.com website. Now, this integration is 37 euros but what it basically means is that you can use it within all of your make.com integrations for life. I'm not sponsored by these guys but this is well, well, well worth the investment for the level of automation that we can get. So what happens is you hit the get access for 37 euros and you get this email. And this email gives you step-by-step instructions on how to add it and also crucially your unique access code that you can use to integrate it with make. You'll then need to come back over to perplexity which by the way is such an epic tool. You click on your username at the bottom and then you come over to API in the top right hand corner. And then we need to add a couple of API credits to our account. I've added five dollars, I've used this a lot and I've not run out at all. It's so, so, so cheap to run but essentially click on the buy credits button and load the account with just a couple of dollars so we can run these automations and use it in the background. Once we've done that, we then come back over to make and essentially the perplexity AI icon will appear for you. Then you'll be able to click it and we're going to click on this one here which is create a chat completion. Now, for the purposes of this video, we're going to leave it as Mistral 7b instruct. That's absolutely fine and then we're going to add an item into it. Now, for the purposes of this tutorial, let's say that you and I want to create content on personal finance and that's our niche and that's what we want to do. And we want to stay abreast of all the latest developments. Where would we go? Let's say we head over to a Reddit forum personal finance and we want to look at the top posts for this month and find topics that we think are interesting and that we think that our customers, our target audience might find interesting. So let's look at this one here for example, right? Please pay close attention to your company's 401k investing schedule. What we're going to do is copy the URL and I've created a test prompt here for perplexity which is hey, you're a summarization expert who can identify the most salient and exciting aspects of any Reddit post and comment section. You'll summarize the URL and also complete further research on this topic to provide data points. Then I give it step-by-step instructions. I basically ask it to give me three outputs. I ask it to create a summarization of the article and then I ask it to summarize the comments to give us some unique perspective from the audience in terms of what's the pulse on how this has been perceived. And then for perplex to do its own insights and research and data points so we can bring some actual facts to the conversation. So all we ever do is change this URL. So in this case we're going to copy all the text. We come over to this essage here. We're going to paste it in and then we assign it the role of user. We click okay and then we're going to run this module and the cool thing about make.com and these no-code automation tools is that you can run this as frequently as you like. So every 50 minutes, every hour, every week, whatever you want. Now we've got this one up here. So let's see what the output is. We come over here onto the output section, come down to choices, plus one, message, role, content. Let's have a look at this. Post summary. I once ignored my company's 401k enrollment assuming it wasn't worth my time but a colleague's unexpected retirement due to health issues made me reconsider. I eventually enrolled blah blah blah. And by the way you can completely edit this prompt and just code, you know, basically cajole it a little bit until you get the perfect output that you like. And then it gives us some perspective telling us what other people have found, like some experiences, the importance of starting early and taking advantage of employer matches, some of advice, maximum contributions. And then we've got data. So we've got some data on it like tax benefits, contributions 401k are made with pre-tax dollars, we're using taxable income, starting early, the earliest starts, blah blah blah. So we've got a summary of the post, we've got some of the comments and we've got some data all from just URL. Perfect. Now the next question is what do we do with that? And I'll just say as well by the way, this could be a Reddit URL, it could be an article, it could be a research paper, whatever you want and you can just add as many specifics as you want here. So let's say we want to do a video on that. What we want to do is add and get an output into Google Docs. So if we just come here, we type in Google Docs verified. Perfect. Then we click on insert a paragraph to a document. Perfect. By drop down my drive, click here and I've said, hey, add this to content ideas, right? So then it has appended text. So what do we want it to be? Well, let's have a look. We look at choices, come down to message, content. Beautiful. So what this will then basically do is take the content from the previous module and put it into that and then we click okay. We've got our document here on content ideas. So we come back over to the whole automation. We hit the run once, perplexes magic in the background. Once it's answered the question for us, it will fire it over to the Google Docs. So if we come back over to the Google Doc, we should see some content. And here we go. Look, post summary, first person, the perspective and the data. And then it will just keep on adding this every single time we do. And we can do really cool, fancy stuff with the formatting just to keep this video simple. I've tried to keep it as basic as possible, but essentially every time it runs automation, it will just add it into this document for us that we can just review and make content from. But then this is where it gets really interesting, where we get these individual GPT agents to do some extra stuff for us. So we first of all want to unlink these. So we come here, we right click and then we hit unlink. And then we want what's called a router. Now a router just lets us say, hey, this is the input and these are several other things that fall off at the same time. So we click on flow control and then we're going to come down here to router. Perfect. So one of the routes is going to be to this Google Docs. So we'll bring that over here. Perfect. And then let's add chat GPT. And there's so many different ways that you can use this, but the one that we're looking for in particular is to message an assistant. So we click on that one Perfect. You'll be prompted here to connect to your OpenAI API. Now here's exactly how you do that. So you can head over to platform.openai.com and in this place here, you'll be able to create API keys for your connection. So you come down on the left hand side to API keys and you can create a new secret key for every use case. So you can actually track your usage based on the automations or if you're doing this for clients, that makes it very easy. Now, standard practice, never share API keys with anybody. They may just run up a huge tab and you may have no idea about it. So what will happen is you'll get a new API key and then when you come back over, you'll enter it here, you'll connect them together and then that will make perfect sense. But then what we need to do next is create some really cool, interesting agents for our new automation. So what we do here is we come down to assistants on the left hand side and you can see, I've got a couple of assistants there already. Making an assistant is unbelievably easy. So what we're going to do is we click on this create button on the top right and let's call the first assistant LinkedIn bot and we'll give them a little bot emoji as well so it's perfect. And then we need some really clear instructions. Now, rather than use a pre-built prompt that I had, what I wanted to do is show you how easy it is to create instructions for this by using a platform like Claude. So if you come up to claude.ai and ask it a question like this. Now with Claude, what I've done here is given it a prompt that is, I think, pretty universal and will help you in almost every scenario. So your role is to create custom instructions for a custom AI bot. Think step by step. The purpose of the bot is to take content and to transform it into a LinkedIn post. The bot will receive a reddit post summary, a summary of the comments on that post and some data. Here is the supporting information to help with instructions and information on the desired tone of voice we would like the AI agent to create the content in. And then all I did here is I basically just went on Google and I did this offline because I thought it'd be a bit quicker, but basically just find the three like seminal pieces of content on how you should create LinkedIn posts, post length, guidance, you know, how we need to structure and think about it. And then what I did is I fed it with my tone of voice or should I say the Monzo tone of voice because they do it, I find, in a way that doesn't really make it feel like it's an AI-led thing. I talked about it before my channel but giving it some instructions on how I want it to sound, you can feed it with your own data. Then we're going to hit enter and then what Claude is going to do for us essentially is give us the prompt to feed into chat GPT agent and then we can experiment with it and change if we need to, but this is going to give us something incredibly detailed. So I've just had a little thing here just to say that please only respond to the content I need for the bot, you are instructing the bot and then Claude here with the second prompt has basically started to give us all the output that we need. Look at this, optimal length, 1300 characters or less, clarity, logical flow, maximum impact, enhanced engagement by blah blah blah, very very very very cool. So then we'll just get that improved a little bit just to make sure it's exactly the kind of thing that we need and again once this is done and complete this isn't something that you have to come back and touch a thousand times over right, you're pretty much just gonna leave it there and this automation is going to run the background you know in perpetuity essentially. Cool, so this is all loading up, fantastic and as you can imagine we'll do the exact same thing for Facebook, we'll do the exact same thing for X. Great, so we copy this and then we're going to come back over to our system and give it the instructions. And what's really cool about this right, we've got the tone and style link down, we've got the post structure, editing and optimization, ensure clarity, logical flow for maximum impact, it's really really good like that. And interestingly look at this, create an attention grabbing headline that captures the main point, open with a compelling hook, surprising fact or brief anecdote. All right, excellent. So then we've got this here which is perfect, the model we're going to go for GPT-4 turbo preview, it does cost a little bit more but overall honestly I actually haven't noticed any of the cost in a significant way. So that's all done and complete and once that's done we come back over to make here, so we click on this one here, we've got the open eye connection done and sorted which is excellent. So the message is going to be the output from the perplexity AI right, so we come down here, let's have a look, let's go to choices, message, content, perfect, great. So then this is here, then we're going to do assistant, we're going to pick our bot which is the LinkedIn bot which we're happy about which is perfect and then we're going to click okay, excellent. And then we're going to hit run the automation and these individual agents I find is so helpful because it's so much easier to have like six, seven, eight, ten agents all doing one really specific thing rather than having one that's trying to do a million different things and they can feed into this in a million different ways as well. Like you can literally have so many assistants running around all doing really, really specific tasks for you that just makes your life so much easier. When I was saying that like there's no limitation to how good this can be, like genuinely it isn't. So let's check out what we've got here then, great. So come down to output, come down to content, plus one, text, value, annotations, great. So look at this, did you know the average 401k balance is over a hundred thousand dollars. What's even remarkable is how much of that comes from free money through employer matching. A recent sharing on a poll with the forum brought to light blah, blah, blah, blah, blah, why it matters, so what's holding you back, let's share some wisdom and it's even given us some things. There are a couple of things that I would change out of this for sure, but like how outrageous is it that we all we ever did is we started with the Reddit URL and then like we've got this thing that's like literally specifically designed now for LinkedIn and then what we can do after that for this particular flow is add in LinkedIn. So we click on LinkedIn, again we'd have to verify that after that in advance, but again it's exactly the same process, create a user text post and then we're here, then we create a connection and then just to touch on pricing briefly, so on make you can get it for free and I've been using the free version for this video, it will cover you for the basics, but as you want to scale up nine dollars a month, sixteen dollars a month, maybe a little bit more. So it's all based on the size of operation but it's very, very, very low cost. The other thing that you'll need of course is going to be with the chat GPT and to get the AI systems within chat GPT you do need chat GPT plus, so that's going to be around twenty dollars a month and then for perplexity we just buy credits as we go but that's five dollars and that lasts quite a while and then you've got the cost of getting the perplexity integration into make. But if you do those things and you get it set up, the applications of this I genuinely think are really significant and I just wanted to run through this one in particular as a kind of for instance of like check this one out and see what you can do, but I'd be really curious to know any ideas that you've got if you want tutorials on getting specific outcomes, let me know down below in the comment section, I can do more of these kinds of videos. I appreciate it's a little bit different than the standard video but I just think this whole thing is just so epic that it warranted a video by itself. But in any case that's it for this one guys, have a great week and see you next time.