https://youtube.com/watch?v=Rxcuk9UVNxg
– I’m giving myself 48 hours to build a completely new business with ChatGPT. (dramatic music) – Mark my words, AI is far more dangerous than nukes. – And I’m going to get it to come up with the business idea, the business name, write the sales page, create the ads and well, do literally everything involved with setting up a new business. (dramatic music) (text whirring) Oh yeah, and it also needs to make at least a few sales ’cause that’s kind of the point of starting a business after, right? I have no idea if this is going to work but there’s only one way to find out.
So let’s dive in. (screen whooshing) (text beeping) I was sitting in my pajamas, sipping on some cold brew coffee and playing around with ChatGPT, getting it to tell me funny jokes, I mean asking it to do serious business things. And I wanted to see just how far I could push this whole ChatGPT thing. So I asked it to replace me entirely as in build a whole new business in just 48 hours.
Was I crazy?
Probably. Was I determined? Absolutely. (screen whooshing) (text beeping) So I’ll be using the power of ChatGPT, a super smart AI language model in order to create an entirely new business from the ground up. And I’ll be treating it kind of like having a virtual business partner that never sleeps and that can come up with ideas and designs and content in seconds.
First, I needed a business idea. So I started a brainstorming session with my new AI BFF and asked it to give me a whole bunch of different creative business ideas. And the suggestions that came back ranged from the mildly amusing to the downright bizarre. Think doggy dating apps and robot hairdressers and AI-powered yoga pants. I’m not even sure how that would work.
Not to mention an eco-friendly toothpaste subscription box and our rent-a-pet service. Finally, after a few hours of back and forth and realizing that I was quickly eating into my available time, ChatGPT and I agreed on an idea, a personalized AI-written poetry service. I’ll be honest, I have my doubts that this is going to work or that I’ll sell even one single poem but only time will tell.
Okay, so we have our idea. Now it’s time to get down to business, literally.
Now I need a name for this business. So ChatGPT, work your magic. Alright, so ChatGPT has come up with a list of different names for our personalized AI-written poetry service, including Rhyme and Reason, Verses Unleashed, The Poetic Machine, SonnetBots, and RhymeCrafters. These are all pretty good actually, but let’s go with Rhyme and Reason as it really captures the essence of our AI-written poetry service. Alright, business idea, check, business name, check.
So next up, we need a website and a way to receive orders and accept payments. (screen whooshing) (text beeping) Now, I didn’t feel like creating an entire e-commerce solution and even modifying and customizing a Shopify theme just wasn’t really in the cards here.
So I decided to keep it super simple and go grassroots using a service called Gumroad, which is kind of perfect for a digital product or service like this. So first off, I’ll ask ChatGPT to write the product description for the service that we’re offering. (lively music) Yip, not bad.
Then a little bit of copy and paste action. And just like that, we’ve got a sales page. That said, it is looking kind of plain. Certainly very boring. So let’s ask ChatGPT to come up with a few sample poems in order to show people what they’re gonna get.
Let’s ask it to write a funny limerick for my fictional brother named Dave. I actually, do have a brother, but his name is not Dave. And maybe an example poem for my grandmother who’s 92 years old. I actually do have a grandmother but she’s not 92 years old.
And finally, maybe a haiku about our recent family trip to Hawaii.
I actually do have a family that recently went to Hawaii so this one’s actually accurate. (lively music) All right, those poems are actually pretty good. So next, let’s see if we can make this whole sales page look just a little bit nicer and a little more professional. Now for any new business, a logo is not mandatory but it’s a nice added touch. ChatGPT doesn’t do images though.
So the best that I’m gonna be able to ask for here is a little bit of direction and strategic guidance and what logo might be best for the service. (lively music) And it looks like ChatGPT has come through again with some pretty good suggestions, including a classic fountain pen or quill symbolizing the art of writing and poetry.
I think that’s probably a pretty safe bet. A circuit board pattern or stylized gear representing the AI technology behind the service. Yeah, I like the idea of a circuit board, maybe some circuit looking thingies.
And third, a parchment scroll or open book to showcase the creative output of the poetry service. Bit of a bonus tip here, but anytime you’re able to use an image that showcases the end state or the result that a customer or client is going to get from your product or service, well, you wanna take that opportunity as it helps to increase sales and conversions. All of this sounds pretty good so I think I’ll find a royalty-free logo, make a few customizations just like this and we’re pretty much good to go. And I think that ChatGPT has also given me enough guidance to find a really good header image for the sales page. So let me head over to Unsplash, find one there that I think fits the bill and we’ll put that on the sales page as well.
Okay, so at this point, we’ve got our business idea, our business name, we’ve got a logo and we’ve got a sales page. Now we need customers. (screen whooshing) (text beeping) So let’s talk pricing. I figured that I should keep things simple and relatively cheap, so I figured five bucks a poem. But wait, in the spirit of hilarity, I figured I’d introduce a pay what you want option.
I mean, what could go wrong, right? Spoiler alert, everything, but we’ll get to that later. First, I needed a way to get traffic to this new business and to this new offer in order to see if it would actually convert. And when it comes to getting fast traffic, the best plan is usually paid advertising and seeing as Facebook ads are one of the quickest and easiest forms of advertising to get started with, I figured we’d start there.
So I hopped back into ChatGPT and asked it to start making me ads, ads that sell my personalized AI-written poetry service in a conversational and kind of funny way.
I figured three or four different Facebook ad variations was probably a pretty safe place to start. So once I had those in hand, hopped into Facebook Ads manager and started putting the pieces together. Now, this is where I feel I need to be very clear about the direction that this video is heading because the second that I press publish on an advertising campaign, things get very real.
This is because advertising costs money. So in most cases, basically all cases other than something like this where you’re essentially testing something completely crazy and ridiculous, you wanna make sure that your entire campaign is completely dialed in before looking at paid advertising.
This means doing due diligence, a little bit of market research, making sure that your offer is clear and compelling. You’ve identified an ideal target market that you wanna sell to, and you’ve worked through a number of different messages in order to properly communicate the solution that you have and why it is the ideal solution for the problem that your market’s going through. But I’m doing none of that, partly because I’m running out of time and partly because I wanted to see just how much ChatGPT could do on its own without me micromanaging all of the details around the campaign.
So it’s time to get posting but I figured the best way to do this was to first ask ChatGPT who it thought the best targeting options would be. (lively music) And it looks like ChatGPT returned a pretty decent targeting list based on interests and demographics, behaviors and connections.
It even offered some custom audiences and some lookalike audiences, which in the case of our brand new business, while these don’t exist yet as we’ve essentially sold nothing to anyone ever, but that’s okay. We can still use all of the interest-based targeting that it’s provided in order to see if this thing’s actually gonna work or not. I’ll also need a Facebook page to run these ads through ’cause I don’t wanna run it through my own personal page, so I’ll just copy and paste a few things over from the sales page that ChatGPT wrote for us, use the same header image and we’re pretty much done there. In the interests of keeping this as simple and hands off as possible, I’m gonna go with the traffic objective as this allows me to not mess around with things like Pixels and cookies and tracking and conversions.
Again, this is definitely not the best plan for most people most of the time, but it’s gonna work here for our crazy poetry service or it’s not gonna work here.
I have no idea. I’ll set up five ad sets, maybe four ads per ad set at a budget of 30 bucks per ad set per day each, and we’ll just let it roll. Fingers crossed, when I log back in tomorrow morning, we’ve got a sale or two or at the very least, we’ve figured out which ads and which audiences are actually going to deliver results and get clicks and engagement. So if ever we did want to theoretically scale up our bot-powered poetry writing service, we’d at least have some pretty solid information to build on. Okay, goodnight.
– [Narrator] The next day. – Okay, so good news, bad news this morning and some more good news at the end I guess. Let’s start with some good news. The ads worked. (people cheering) Sort of.
(audience booing) I mean the ads got approved, got shown to the target audience and even managed to get a few likes and clicks and comments.
Now, most of the comments were negative. Almost none of the clicks converted. More on that in just a minute. And at least from an ROI or return on investment perspective, things could not have gone worse.
But we did manage to make one single sale but that’s okay because there are some seriously valuable lessons to be learned here about how we could have made this entire thing a whole lot better. So let’s talk about that now. (screen whooshing) (tranquil music) First off, a little bit more thinking and planning and strategy would’ve gone a long way here. Had I run the idea for a personalized AI-written poetry service through my own marketing framework instead of getting excited and quickly handing the whole thing over to ChatGPT, I would’ve quickly realized that this offer or idea didn’t really check any of the boxes that I look for when I’m launching or promoting a business, like market demand or existing competition, proven track records, access to a large enough customer base or scaling potential.
The only thing it kind of hit on was the novelty and uniqueness factors of using artificial intelligence.
But when paired with a service like poetry, it’s again pretty easy to see how this falls flat. (screen whooshing) The sales page was terrible. Again, in my rush to get this whole thing up and launched in 48 hours, I tried to put together the most simple, minimal page in existence but that meant foregoing important conversion-boosting tools like testimonials and social proof and ratings and reviews and guarantees and even better, a more clear description about what was actually being sold and what people were actually going to get. Ideally, I would’ve liked to see the page convert at at least 2%, meaning for the 141 sales page views that we got, would’ve made anywhere from three to four sales, not one. (screen whooshing) The Facebook ads did what they were supposed to and drove traffic to the sales page.
Not only that, but they also gave us valuable insight into which ads worked, which didn’t. And when I sort the ads by which ones delivered the most link clicks, I can also see a pretty clear pattern about the style of image that works best with this offer and audience.
It’s this image, by the way, not to mention all of the other metrics associated with the campaign actually look pretty good. I mean, we got a CPM cost per 1,000 impressions of around eight bucks, managed to get a CTR, a click-through rate of 2.75%.
And then what I look for is a link click-through rate of half the click-through rate, which would be 1.375% but we managed to get 1.84%, which means that not only were people clicking on the ad to read more and click the like button or leave a super friendly comment but that more than half of these people were then also clicking on the link to take them to the sales page to learn more about the offer, which means the biggest problem with this whole thing is that this was just a genuinely terrible offer. (screen whooshing) A genuinely terrible offer that nobody wanted, except for this guy.
And I just realized I now have to write a poem for him.
Now, sure, you could spend time and money and energy trying to squeeze out a sale or two, micromanaging every step of the process but you’d be much better off fixing or ideally completely replacing this offer with something that people actually wanted. Advertising is an accelerator, not an initiator, which means that it’ll take whatever you have, good or bad, and then make it win faster or lose faster by putting it in front of more people. (screen whooshing) ChatGPT gave me exactly what I asked it to do. I asked it to come up with business ideas and it did. I asked it to write me a sales page, and it did.
I asked it to create Facebook ads and it did but it was up to me to audit those answers, to review them and to act like a filter to make sure that the idea that I was about to run with was solid.
And AI can’t do that yet, but to show you what AI can do and what things look like when they actually work, then you’re definitely gonna wanna check out this video right here where I’m gonna show you step by step how I got ChatGPT to build me an entire marketing campaign. So make sure to check that out now and I’ll see you in there in just a second..