3 Jobs GPT-4o Can Do For You Right Now

Okay, look, so OpenAI released a brand new model that is more human than anything they had before. Oh, majestic potato. And if I look at the comments section of my recent videos, it becomes apparent that there's a whole new group of people that are freaking out because this is going to replace all of our jobs. And I'm creating this video to address that. But I'll do it in my very own way because I'll be looking at these jobs and I'll be actually showing you how to do the techniques that supposedly replace the jobs. Because when I talk to these people that are worried in real life, I often find that they don't have full clarity on what is possible today with this technology. Or there's the second camp that is worried about a different timescale. They're not talking about jobs being replaced today, they're talking about it over a 5 to 10 year timescale. And you know, I think that concern is perfectly valid. But that's not what we do on this channel. We talk about how to use this technology today to improve your life. So here's a very practical look at what you can do with AI today in these three jobs. And my hope is that it both informs but also educates people how this is actually going to impact these jobs, rather than this unmanaged anxiety of, oh, it's going to replace all jobs. What are we going to do? Nah, let's rather go case by case and learn something along the way. Let's start talking about data analysts because a lot of people claim that with the Data Analysis Tool, aka Code Interpreter, inside of both ChatGPT 4 and now 4.0, a lot of the work of data analysts has become obsolete because people can do this themselves. And I don't dare to make estimates of how many of these jobs are actually going to be effective because what I am here to do is to show you what's possible, possible today. Matter of fact, I really want to talk about the data analysis tool because it improved so much and they didn't even make a big deal out of it. So I'll head on over to chatgpt.com switch to ChatGPT 4.0, the brand new model. And what we'll be using here is the text capability and the data analysis tool, also referred to as the code interpreter. I still call it the code interpreter because that was the name when it first came out. But here's the deal. There's so many new features surrounding this. Starting with when you click on this attach tool, they updated it and now you can connect directly to your Google Drive or, or upload from your computer. Not just upload from your computer, you get this fantastic choice for ease of use. And now I can browse my Google Drive. And the thing that might not be obvious from the outside is the fact that you can actually open Google Sheets and Google Docs, including Word and Excel sheets from here. So that makes some of these workflows super fluent. That's a simple speed improvement when you're doing data analysis. But I'm just going to do it the good old fashioned way. I'm going to upload from my computer. I have a little data set here that I compiled for this tutorial. It's basically some survey answers from members that recently joined the free area for community. And it talks about what they want to learn about in relation to ChatGPT, a relatively simple data set in a CSV format like you might have it. You can export every single piece of analytics or surveys or Excel sheets would behave very similarly as this. And this is where we get back to this tool replacing data analysts. Because one might think that, hey, this is a sophisticated tool inside of AI interface. Most people won't know how to use that. Well, what if I told you that the only thing you really need to know to get started here and to become sort of a data analyst yourself is one question. It's this. What could I do here with data analysis? And then you just hit enter and you don't even worry about it. And first of all, you'll see this new ChatGPT feature in here where it's actually an interactive table. So you can do various things here like expand it and then you get this brand new side view. Or you could just go ahead and selectively pick one of these rows and then you will be prompting with only this row as the context. This is very powerful if you know what you're doing. You really get to be selective about the context here. But back to this data analysis, because look what it did here. It gave me examples of various things that I could do and I might have no previous knowledge of this. Matter of fact, a few months ago I've never even done an exploratory data analysis. But what I can do now is I can just review this and I could say something like do 1 and 3 and then hit enter. And now it's using the data analysis tool and I don't even need to have a clue what I'm talking about. And then look at that, I just created this one. Okay, not great. Okay, no problem. I'll just re prompted it to use data analysis todo8 and hit send on this. And here's my favorite part. This 4O model made it more conversational and improved the style. But also in the way it handles the data analysis tool, it's actually way improved. Look at it. First pre processing the data and then using the data analysis tool, which by the way is so fast. Now they said that this model is twice as fast, but honestly, just from my intuition, the data analysis tool, it feels five to six, six times faster than before, making it actually usable. Because look, before I even finished the sentence, it finished analysis. Here I have a distribution of all the unique goals. I don't know how useful this is, but it certainly looks very interesting. And these are the new interactive graphs in here. So you could change colors of specific data points, you could download it, you could change the views if they don't fit you like so. And here's the nice thing, because I started with a beginner prompt like what could I do here with data analysis? It is aware of the fact that that I'm not even sure what to do. So it gives me little suggestions like this down here. So you know, I could just follow up and say create a heat map or how about a network graph? And this is the whole point. This tool got way better. I barely know anything about data analysis, but I'm getting some useful insights here. Matter of fact, I had a conversation with this exact data set, just the full one, for about two hours before this. And I was just blown away by how many insights I was able to extract from this data without knowing anything about the subject. Okay, this looks a little messy. I could follow up with something like develop some actionable insights from the network graph. And you can see that these prompts are very simple. This doesn't take much skill. Nevertheless, it manages to give me actionable insights that are data driven. And I'll end on the point that this primitive prompting demo of mine here is just the tip of the spear because some people are developing specific tools like Julius AI that added support for R based analysis. If you're not familiar, R is one of the most widely used languages used for statistical analysis. And all of this can be done with AI now. So the point is not that maybe me plus ChatGPT are going to replace a lot of data analysts, but a very competent data analyst with tools like this is literally going to skyrocket his productivity and people like me are going to be able to get a lot of day to day Tasks like this done by themselves. And that is the status today. Because with the speed improvements, interactive tables and graphs, and the new, improved model, and the fact that soon I'll be running around with my phone pointing at my graphs and clicking through my analytics and talking to it live, simply, judging from what I can see and do, I think people that are worried about the demand for this job diminishing have a point. Now, let's talk about the next one, which is customer support agents. This is the big, obvious one that has been a topic ever since ChatGPT became a topic. Oh, my God. Okay, it's happening, everybody. Sitcoms. And that's because the bar is very low, right? When you call a bank, usually you'll be forwarded to this robot that is like, if you have an issue with your bank account, press 1. If you have an issue with your credit card, press 2 and so on. Those things take a lot of time. Sometimes you're not able to find the thing that you're looking for because their preconceived notion of what a problem is might be different from what you're actually facing. Plus, then it puts you into this queue of elevator music where sometimes you sit in there for 20 minutes. Well, AI fixes that. And this is the one where I would argue, sure, a lot of jobs are going to be replaced in this segment, but this one in particular is ripe for disruption because all of these companies going for these automated phone systems, they're doing that at the benefit of their bottom line and at the cost to the customer. And I think it's about to get a whole lot better. I'm going to show you a concrete technique how you can actually improve the output of your customer support agent that you might build with ChatGPT. But first, I just want to point towards this customer service proof of concept. We already talked about this on the channel, but he basically uses two phones with Chat GPT4O and the new voice assistant that hasn't shipped yet as of today, May 19th. When did Joe receive new iPhone? IPhone was delivered two days ago. Cool. Could you share the order number with me? Of course. It's 1-029-3849. And then they resolve a customer support ticket like so. Now, for the everyday user, there's a few issues here, right? First of all, you could prompt it with something simple, like the simplest prompt, and I'll just add what they're complaining about. Let me show you. If you just do this, it's probably not good enough. And a lot of people will argue that, hey, a bot might be better. But I'm here to clarify something that is not so obvious to people that are getting stuck started with these tools, and that is the fact that you can really flesh out the context that you put in here. So it's not just prompt engineering tricks that you can include in your prompts. It's really just data about the subject that you're interacting with. And in this case, there's two people in this interaction. It's one, the customer support agent that works for your company, and two, the customer. So now one of the first questions that I would ask myself in order to improve this prompt would be, what info can I give ChatGPT 4.0 so it can do a better job? It's trying, but I'm not giving it a whole lot to work with. And if you've been watching the channel, you'll know that I'm a massive proponent of using custom instructions in pretty much everything that I do. And the main way I use them is actually these presets that give it a little bit of info by yourself. You might also know that usually you have this default which basically states that I'm a YouTube content creator. It states some of my interests, my values, my goals, my learning style. And in a second I'll show you where to find a bunch of presets like this that you can use for yourself, completely for free, by the way. But all I'll do is put this email in parenthesis so it's separated. And then I'll say here's my personal context and I'll just copy paste these custom instructions here. Okay, defining all these things about myself. And I'll just say send. And now look at the result. Transform. It all of a sudden talks from my perspective and it follows some of my values of customer driven iteration and improvement by asking for feedback, being helpful, writing in my tone a bit. That's not Perfect. It's still ChatGPT and we didn't employ any advanced techniques on how to do that here. And it even promotes the newsletter, which I wouldn't want it to do. Then I would customize this part of the program prompt here. Right, but here's the point. Just by copy pasting this, I went from a generalized email to a customized email. And you have to consider that I gave it the information on me. But the number one thing a customer support agent would want would be the information on the customer. So think about this. If you're a small service business and you only serve a few people, you could have Profiles like this set up for each customer and then whatever you might be doing with ChatGPT, it might not be customer support. Everything would be enhanced by these custom instructions. So let me go ahead and say the same thing, but other way around. Here is information on my customer and I'll send this and watch how to rewrite the email in a way where it tries to appeal to a YouTube creator that has a problem with the template. Instead of me being the YouTube creator helping somebody who can't find the link and look at that in every paragraph it references some of the context. It says understand the importance of having all your resources at your fingertips. It understands the values and the needs of the customer. Now look, I'm not advocating for you going out and replying to your customers like this. This is a starting point and there's a lot of prompting that we could do here to modify the structure of this. We could include multi shot prompting where I give it five examples of cloud customer support emails that I've written before and then it follows a template. Maybe I want it more concise. We could give it a list of different notion templates we have with the links and then it will insert those too. We could take all of this and build an automation on make on Zapier that looks at every single email that comes into our inbox. Uses this prompt with the modification that I mentioned to automatically create a draft for every incoming email. Plus it can consider the context for your various customers and for yourself at point this the same time. And all of a sudden you have a very simple customer support system that actually works and all it takes is a functional prompt, a simple automation. And we're already at a point where this is so much more meaningful than setting up one of these automated robots or in many cases paying a person that does this full time. Now given these are just text based queries that we're talking about here, but with the voice assistant clearly also some of the voice based ones will be able to use the results of your prompting in an interactive way. Voice based interface with no latency. Plus you can record the calls, transcribe them, improve your prompt, improve the context over time. So it only becomes better and better. Now a second ago I promised you that I would show you where to find presets like this and it's actually inside of this free file that I provide when you sign up to my newsletter. It's really very simple. Sign up here once a week you'll receive an update on all the things that are happening in the AI Space what tools, prompts and workflows you could be using. It's very action and use case oriented. Plus you'll get this file for free. And if you go to this first section here, ChatGPT assistants, you will find this tab called called Custom Instructions. And basically here are the custom instructions for 10 different jobs. And you could just take these, manipulate these and use them as the personalized context, if relevant, for various prompts. And then if you enjoy that and would like to have a database of a thousand jobs, we sell that as a separate product that I call the Business Blueprint. But all the basics are here for free. That's just relevant if having 100 times more presets like this would be useful to you. Okay, one last note about this use case is that this gets really crazy if you give it the right type of context. So imagine if somebody had your browsing history, then they can totally use that to reply to the email in a way that would resonate with you on a level that is way deeper than a general understanding of who you are and what you like to hear about. It's like those creepy ads that pop up on your phone if you talk about something in your apartment and all of a sudden you get an ad exactly about that. Well, imagine that. But instead of it, you have a human like voice that is speaking to you and trying to convince you of something. And then you recognize that it obviously has some intimate details about something that you didn't expect it to have details on. Oh yeah, that's going to get a bit weird. But now let me talk about the third one on my list, and that is social media managers. And I called this out pretty early and this is actually the industry that I worked in. I had a video production company before this channel and let me tell you, even back then I thought to myself that so many of these social media managers are just ripe for disruption. Obviously there's a wide spectrum, but in my experience it's a super easy to do job that you can really quickly grow into. And most of it is just connecting several apps and doing some basic copywriting and maybe creating some simple visuals to support your messages. These are all the things that AI is really, really good at. But it's also the reason why this hasn't really happened yet. At scale at least, all of the points that I just named were notorious weaknesses of AIs. The copywriting sounded very chatgpt, like if you have any standards not usable out of the box. The character consistency on duels like Dall E has been virtually non existent. Text was ungeneratable. Therefore you couldn't even get around using something like Canva because you always needed to add the text on top of and you couldn't create custom stuff. There was no function where you could just upload one picture and then play with that picture as if you knew Photoshop. Well, as you might know, all of this just changed. If you look at the blog post, all of these examples of the yet to be released new image editing tool, they solve this. You have character consistency across Gary the Robot here. So now we can tell stories, create storyboards, create Instagram posts of the same character. But this is just one component to the social media manager cocktail because the other one is really the text to font ability. Now it can do full fonts, it can write anything flawlessly. So now you really don't need Canva to add text on top. And then the third one is this ability to one shot fine tune. What that means is you give it one example and you create a fine tuned version of the model that can recreate it in other scenarios. So you give it one picture, it will recreate that into multiple variations of it, all of them will be consistent and then you can even do text. These are all the reasons this hasn't been used for social media content creation yet. And with this new imaging tool, they're all solved in one sweep. So that's concerning. But then you pair that with the fact that Google released Gemini advance here and bear with me a second before you hop in the comment section, you're like, oh, Google no, no, Google no good. Listen, this is the biggest context window we have yet. So you can do some things that aren't thinkable in other contexts. And some of this is really, really good for social media content creation. Now here's a quick example. OpenAI recently announced that they're going to be partnering with Reddit and pulling in some of their Data to keep ChatGPT up to date. This is similar to what GROK is doing with xdata. But while that is all well and good, with large language models like Gemini that have a context window of a million tokens, you might not even need this because you can do it by yourself today. And this is what I'm going to show you here. You could just go on Reddit. Here's a random post on the ChatGPT subreddit and what I'll do is I'll simply do this. Everybody can do this. You don't even need to web scrape. I'm just Going to copy this entire comment section. And yet again, look at all these simple prompts. I want you to give me ideas for Instagram posts based on this Reddit post and comment section and then I just paste it in and I'll do one more thing which is enable my custom instructions so it knows about my personal context. Who am I? What am I trying to achieve? I always like to have this on when creating something like this little ideation session here. And you might see this is again one of these cases where no, we're not at the point where a swarm of agents just replaces a full social media department in a company, but we are at the point where one social media manager can seriously up their game with these two tools and then that state is just one step away from this agentic future where it's clearly going. Because right now I could package everything I showed you in this video in three different GPTs, then flesh them out a little bit, add some prompting tricks, more context. Maybe given some tooling in terms of actions, maybe I wanted to access my notion database, save things in there and then all it takes is one upgrade from OpenAI, which is like now you get to create one boss GPT and give it free GPTs to work with. And all of a sudden my social media manager is working with my data analysis and my customer support agent, which are all plugged into the different apps that I use. All it takes is a little bit of reasoning and all of a sudden these post ideas could be a reality and could be scheduled on my Instagram with visuals and captions created for them. And that's my important point right now. It does take these manual steps. You can automate a lot of it, but not everything. Except if you're Klarna and you have a very clear cut use case. But these examples are rare. Most people will be using this. And that's why I want to show you one more thing and that is kind of the ultimate beginner use case, in my opinion, and that is answering the question, okay, I'm a social media manager or what can I do with these tools? Actually, this was the number one question that I received with various jobs and various pieces of context, of course. So last year we set out to create a comprehensive solution and that's the business blueprint. In the free template you get 10 examples like this, but here we have over a thousand and all I need to do is go and say social media and here we have different jobs. So I'll go with the default social media marketer and here's my answer to this question of what do I do if I have Job xyz? My answer is learn how to use these tools. Learn how to define your personal context either by yourself or you have it in this database. And then familiarize yourself with with the prompt generator that we also provide for free in the first template. Now here they're customized to every single one of these jobs. So if I copy the social media prompt generator, all I have to do is paste it in here, hit enter, and now it's going to generate 30 ChatGPT use cases for me. And look, they're always going to be different, they're always going to be a little random and they all depend upon what you include in this prompt. And then you have base instructions for all the different things that you could be doing with AI today. You can expand this with your personal context, you can add other prompting tricks, you can add multi shot prompting, you can add an interaction sequence, you can make this personalized to your writing style. But this is where it starts. These are the little seeds that trees will grow from over time when it comes to doing a specific job with AI. These are all the different use cases. And that's what this channel is about. That's showing you what's possible today, how you can apply to your very own life. And all of that will hopefully help you adapt to this new technological revolution happening right in front of our eyes. So I hope this video managed to alleviate some anxiety from some people by showing you what's actually possible. I hope it was helpful to some others with some tips and tricks. And if you're curious about any single one of the techniques that I brought up in this video, we teach all of this in the AI Advantage community, which is paid, but you get instructional guides and lectures on all of these topics, plus my teams and the community support while you're learning. And that's for the whole new setup for the brand on the channel. We show you what's new and why it matters and then the products teach you how to use it or save you time while using it. Simple as that. And I will see you in the next video.