ChatGPT Automation: Why It's Harder Than You Think

LARRUZ breaks down why AI automation agencies sound easy but aren't. You need GPT-4, Zapier, and chatbot platforms, plus prompt engineering and basic Python skills. The field moves so fast that solutions built today might be obsolete in months.
Chapters
The AI space is evolving and right now everybody in their grandma is talking about AI automation agencies. But I'm here to tell you that it's not the opportunity that everyone makes it out to be, because I see most people comparing this concept of automation and AI as a service to social media marketing agencies. And while there are similarities, they're very different. So today we'll be diving into those differences and what you actually need to deliver on the promise of an AI automation service. And look, I'll be honest in this one, the gloves are off. And it feels like I have to add my 2 cents here because I've been living and briefing this low end of the market in the past seven months. And what I mean by low end is the tools that are available to consumers, ChatGPT, prompting automations, other AI tools, basically all the common tools. Whereas the high end of the market would be enterprise grade solutions that this entire AI automation agency topic isn't even touching on. So with that being said, let's have an honest look at what an AI automation agency is supposed to do. And if it's really as easy as many people put it out to be, spoiler, it's not. So we'll go for this video by comparing automation agencies to social media marketing agencies. You're probably familiar, but that is the popular business model on the Internet of finding clients that you can run ads for most of the time in return for a cut of the results plus a retainer. And here on the other hand, with the AI automation agency, which is all about streamlining processes and reducing costs, either by reducing staffing costs or helping companies increase productivity by getting more out of the people they already have. Nice. Now if you compare the two of these, the main difference that most people hit on right away, and the main selling point is, is the fact that there's a dramatic difference in demand right now. Social media agencies have been around for a while and the market is saturated as heck. Companies that realize that it's valuable already have their internal teams and many founders have mastered the skill of social media ads. They don't need an agency to supplement that. Plus this massive hype wave of people across the Internet jumping on that trend and offering their services creates a situation today where there's a massive amount of supply and a decreasing amount of demand. In other words, it's really hard to sell to clients. But on the other hand, with the AI automation agency, it's an entirely different story because as you might have noticed, AI has become a thing to the mainstream and absolutely everybody is aware of AI at this point after the hype around ChatGPT and MidJourney. And people are very emotional when it comes to this topic, right? They fear the change that is in the air, right? The job market is going to drastically change, which isn't just a threat for employees, but also for founders. Competitors might use these technologies more effectively and outcompete their companies. Plus, if you mix that with all the AI fear mongering and all of that results in a state of mind where you want to take action to get rid of that fear. But there's also curiosity, right? They probably tried ChatGPT and they probably had some decent results, but didn't really find a way how to improve their business with some of the prompts they saw on the Internet. Now that is the demand side of the equation. On the supply side, you have not a lot of people offering this because these AI automation agencies for the low end of the market, right? For the small to medium sized businesses weren't really a thing a few months ago. Not at scale that is. But the reason this model has been popping up more and more over the past few weeks is because the introduction of GPT4 and now the wide access of the GPT4 API allows even for people that are not very skilled at prompting to generate decent results that they can keep selling to clients. And that's just the automation side. There's also the chatbot side to all of this and that has seen a lot of technical developments over the past few months too. That is Sensei Igor for you. Apologies Sensei Igor. I will make sure to address you correctly. And obviously those cheap chatbots are all built on GPT. So with larger context windows and more capable models, you're all of a sudden getting results that are actually useful to businesses. But not a lot of people know how to provide those services, right? So you get the situation where there's a lot of demand because people are scared and curious. And then on the other hand, there's not a lot of supply because the basic model of providing chatbots and using zapier plus custom prompts to help people streamline their business at a relatively affordable cost just wasn't a thing in early 2023. At least not with this level of quality, ease of implementation and a relatively affordable price for some of these low end chatbots. And what I just laid out there is the main reason a lot of people are pushing the concept of the AI automation agency. Because there's a lot of demand, low supply, which makes that concept very easy to sell, which is exactly what a lot of these people are looking for. They're looking for efficient way to make money. And there's nothing wrong with that as long as you can follow up on the promises that you make to your clients. And that brings me straight into the next topic, which are the tools that you need to complete these jobs. Because with the social media marketing agencies it was quite straightforward. You had your ads manager, you had to learn a bit about analytics, but generating some traffic and some sales when the client presents you with a budget to spend is not rocket science. And the tools were pretty intuitive. Facebook ads and Google Ads were always built in a way that they were relatively approachable. You might need two days to get up and running, but you don't need to know some super specific software or coding language to engage in that business. On the other hand, for the AI automation agency, it's a different story. You, you actually do need specific tools that can do a lot and you need to learn the capabilities. Now this doesn't mean you can't use presets in something like bot press or use pre made zaps inside of Zapier to get some tasks done. And even disregarding the quality of your system prompts and prompts that you feed to these bots to get productive outputs, something like zapier can be very, very deep if you want it to. Interacting with webhooks conditional logic. Over 5,000 apps that integrate together and can break in different ways when things are not set up correctly. Look, they did their very best to make it simple, but it's a complex task. It's not. Now producing three different ad creatives and testing them with $10 each and then putting more money on the winner and repeating that process for the next client. But look, maybe this part I'm exaggerating and it's actually easy to build some basic bots with something like bot press. And if you're completely new to Zapier, also in two days you will be able to get some results that are actually representable. The problem is that the prompts you will be using with the ChatGPT integration will probably not yield the results that you would ideally want. And that brings me to the next topic, which are the skills that you need for this because obviously in both cases, as we talked about, you need sales skills, but the sales are going to be way easier for the AI automation agency as you just have this misalignment and supply and demand. But beyond sales, which every agency model has in common, for the AI automation agency, you actually need prompting skills and some basic technical skills. And this is what I have an issue with. With a lot of the content I've seen on the Internet. People keep telling you that it's very beginner friendly and you can just go in and you can just copy their Zapier templates. But here's the issue. When you start talking to clients and working with clients and every situation will be different. It is what it is. Especially with automations. Even when we talk about the basic social media machine where you come in with let's say a blog post and then you reformat it to email, tweet, threads and other type of written social media content, that whole automation will really depend upon your input. In what format is that blog post written? Or maybe is it even a blog post? Maybe the source for the information is some news article or maybe the client wants to fill out a custom form every single time they have a new piece of internal information and they want to create social media content from that. Well, if you dive into Zapier you might be able to figure all that out. But. But will you be able to figure out the prompting part of that to get the tone, voice, sentence structure, content exactly right for the platform that you will be publishing that content to? Well, let me tell you, I've been actually taking on some client jobs next to the YouTube channel over the past few months and it's not that simple. Now yes, if you stick to that one to $2,000 range and you just deliver basic chatbots, that is quite simple but that is also not an opportunity that is very sustainable. All it takes is Microsoft launching a brand new Microsoft Teams feature that is a custom bot that is based on all your company's knowledge and that basic service is purchased probably going to go away because people are going to be able to click to buttons and it's going to happen. Now look, at this point in time it's definitely valuable. For example, we built a bot for fun inside of the course members area here in my discord. And if you say/GPT you can ask it any question about the course and it's going to provide you information from the video transcripts plus a link to that specific video. Let's say how can I stylize my text inside of ChatGPT? And there you go, you have an answer and you can actually click the video which leads you straight to the correct video inside of the course course platform. Now I think it's needless to say that something like this can be very useful for companies Right. If you pack all of the HR department's knowledge and FAQs into a chatbot and you use it for either onboarding or to answer employees questions, well, you just took a lot of work away from your busy HR manager. Same thing with some of the marketing automations that I just outlined, right? You can really help the social media marketing team with their tasks. You provide them with more options, but those things are not that easy to build anymore. As opposed to the social media model where you pick a target audience, get give it some creatives and run some tests and see how things go and then move on from there. And that brings me to the last point, which is the industry that these services are located within, the social media industry has developed quite fast, right? And since Facebook ads started getting really popular back in 2014, it took five to seven years for the market to get kind of saturated, Right? But when you look at the timeline, sure, there were some drastic changes with Facebook changing their approach to organic traffic on business pages, and they keep changing that ad manager around all the time. But a lot of the things that held five years ago still still work today. And it's essentially the same platform where you go around and click and upload things and pay some money to promote the content. While it's definitely not a slowly developing industry, if you compare that to AI, it's like a race car passing a kid on a bicycle. And the point that I'm trying to make here is that the things that go today might not work tomorrow. I mean, look, seven months ago we got GPT 3.5 with the release of ChatGPT three and a half months ago we got GPT 4. And look at the development of all the AI art generators. Half a year ago we weren't able to process, properly generate anatomically correct human beings that are photorealistic. With MidJourney 5.2, even beginners can do that with two simple keywords. And that's why I think that only the people that can adjust to these changes at a fast rate and maybe have a bit of a technical background, maybe know at least the basics of Python to implement some custom solutions and to understand the logic that goes into building some of these automations, or hey, what about this? Maybe having a little bit of business experience. So the companies that you will be serving with your own automation and chatbot services will actually get some value from you coming in there. But that can only happen if you understand how their company works. Because an AI automation agency is not in the business of generating new customers. It's in the business of making a company better from within. And at this point I also want to say that people like lime on YouTube absolutely know what they're doing because they have that technical background. But I just want all of you to be a little more vigilant when you look at some of these videos where obviously the main reason they're talking about this topic is is because it's easy to sell. Yes, that's fair and there's real value in these services. But consider that a new model might be released tomorrow and these agencies will have to readjust to serve their clients well. And I don't think that the people that come in only because it's an easy sell will be able to thrive in that environment. But hey, look, at the end of the day, results speak louder than words. And as you might have heard already, AI Advantage is also offering agency services from here on out. We have a client waitlist that you can sign up for now. And what we're providing here is a little different from the market because as you might have know, I've been building a team across the last seven months and I put together state of the art prompt engineers with experienced technicians that can deliver custom solutions that go beyond what you might be able to learn in Zapier over a weekend. But again, the results should speak for themselves. We're already working on some case studies that will be posted to this channel, but for now, I'm curious to hear what you have to say. Leave a comment below with your thoughts. Do you agree or disagree on some of the points I made here and why? And if you want to learn more about how to deliver services like this, I strongly recommend you check out this video that teaches you how to get to the next level with prompt engineering. Because simply copy pasting what you saw on Twitter is not going to cut it when you're going to approach some of these custom solutions. All right, that's it for now. Back to work.










