DON'T Use Claude AI for Business Until You Watch This

So MIT just released a study on AI in businesses that found that something like 95% of companies trying to use AI are failing at it. McKinsey did a similar study, looked at almost 2000 companies, and found that only 6% are actually crushing it with AI. So before you go and use cloud AI for your business, you need to watch this, because here's the part that nobody is telling you. It's not that AI doesn't work. It's that almost everybody is using it completely wrong. And the fix is the opposite of what you think. It's not fancier, it's not more complicated. In most situations, it's actually simpler. And companies are lighting something between $30 and $40 billion on fire trying to bolt AI onto their businesses and basically getting nothing back. Meanwhile, regular people running one-person operations are quietly using the exact same tools to print money. Why? Because the big guys overcomplicate everything. I like to call this midwitting it. And you don't have to. And I'm gonna be giving you the five rules that separate the 5% who win from the 95% who waste their time. And then I'm gonna show you the single highest leveraged place to point all of it, the one that's made me millions. Now, real quick before you dive in, if you appreciate me making this type of content and you want to see me make more of it in the future, then let me know by gently tapping that like button. And let's jump into it right now. Rule number one is to start stupid simple. And this might be the most important one on the whole list. So do not skip it. So here is a law of the universe. Every complex system that actually works evolved from a simple system that worked first. And that is not just me being clever. That's an actual principle. It's called Gall's law. And it was written down by a guy named John Gall. And it goes like this. A complex system that works is always built up from a simple system that worked. And the flip side of this is a complex system designed from scratch basically never works. You can't even patch it into working. You have to start simple. So think about it like building, right? You don't start construction on the 50th floor. You can't. You pour a simple solid foundation, then floor one, then floor two, and the skyscraper earns its way up. If you try to start at the top, the whole thing collapses before it even exists. And it's the same with AI. Everybody wants to build the giant automated 50 floor AI machine on day one. But Shane, won't the simple version make me look like an amateur? No. The amateur is the one who spends three months building a monster that never ships. And then by the time it finally does ship, there's some other AI out there that's even better. And so you gotta rebuild it again. And the data backs this up hard, right? Remember that MIT study? Companies that tried to build their own custom AI from scratch failed about three times more often than the ones who started simple and just used an existing tool. So building the complicated thing from zero is three times more likely to blow up in your face. That is Gall's law operating in real time with a very real price tag. And if you don't believe me, just look at probably the smartest entrepreneur in the world, Elon Musk. There's been many times where instead of trying to automate something, he literally will just have real human beings do the thing on the production line. Now that's about as simple a system as you could possibly have, but he would rather have a real human being do it slightly less efficiently than automations do it, but do it really badly. So here's how to start. Pick the simplest possible version of what you want, right? Like embarrassingly simple. Open up Cloud and instead of building some elaborate system, just do one small useful thing today, right? Break up the entire thing that you're trying to do into its component pieces, and then figure out how to do each of those component pieces really, really well. One email, one outline, one task. Get that working first. You earn the complex system by nailing the simple one, right? Do not pour the skyscraper before the foundation. It almost never works out well. And we're gonna get into exactly how to design those systems later on with examples. With that being said, let's move on to rule number two, which is actually a natural partner to rule number one, which is the simplest answer is almost always the right one. Now there's a famous principle for this one too. It's called Occam's razor. And basically it says when you've got two solutions, the simpler one is almost always the correct one. Do not add complexity that you don't need. And when it comes to AI, the boring simple use is almost always the one that actually makes you money, right? It's like ordering at a restaurant. You ever go to one of those places with a 12 page menu and a hundred different options, and then the food ends up being mediocre across the board. Now compare that to a tiny spot that does one thing really well. Like for instance, maybe they make really good breakfast tacos, and maybe they have five different options for really good breakfast tacos, and they're all incredible. The simple menu is gonna win every time. So focus does the same thing for you when it comes to AI. And again, the data agrees. The lead researcher on the MIT study said that the companies absolutely winning with AI, including tiny startups that went from zero to $20 million in revenue in a single year, they didn't do a hundred things. They picked one pain point, one, and they executed it really, really well. That's it. So simple beats complicated basically every time. So here's how to start with doing this, right? Write down the one biggest, most annoying, most repetitive problem in your business or your content creation process right now, just one, then point Claude at that and ignore everything else for now. A really good way of doing this is doing what I like to call a yap session where you just put on voice, and then you just yap to it and talk to it about your problem. And then what I really like to do is have Claude interview me until it has enough information, data, and context to actually be able to solve the problem. So solve your single most expensive problem first because the simple focus swing is the one that typically connects. Hey, quick break. I'm gonna be doing a live training this week on treating YouTube like a business. So this is gonna work for people who have online businesses, in-person businesses, or just people who want to treat YouTube like a business from the beginning. And it's a completely free training, no strings attached. In fact, I'm giving away even more bonuses at the training. So just as an example, I'm gonna be giving away my Niche Validator. This is gonna be available both for ChatGBT as well as Claude, which is a super valuable piece of software where you can finally figure out what the best niche for you is. So do not miss out on this training. Make sure you sign up for it down in the description and the pinned comment below because you only get it if you join the training. So if you don't join now, you might miss it. But that being said, in the workshop, you'll get to meet me and you'll get to ask me questions directly. So I look forward to seeing you in there. So click the link in the description and the pinned comment below. Put it in your calendar. And if for whatever reason you missed out on or you weren't able to attend, make sure you still click that link because we might be having workshops in the future as well and you'll be the first to know about it. So yeah, hope to see you there. And now back to your regularly scheduled content. Okay, so rule number three is build it like a Lego. So this is the one that saves you from that rebuild it every week nightmare that I talked about earlier. You wanna make your AI setup anti-fragile. And what that means is simple. You build it so that when something changes, like a new update drops, a better tool comes out, you don't have to tear the whole thing down and start over from scratch. You just swap one piece out. So this is like thinking in systems. And I'll actually put a picture up on the screen of me literally drawing out my entire content creation process on a whiteboard. And I do it with a flow chart, AKA I draw it out like a system. And if you make this system fragile, a fragile system is going to break when the world changes. An anti-fragile system just swaps one part and it keeps going right on. Think about it like a Lego. If you build something out of Lego and you wanna change one part, you don't smash the entire thing and start from zero. You just pop out the one brick that you don't want and you snap a new one in. Everything else stays exactly where it is. So that's how you want to build with AI. You wanna build it like a system, modular, in pieces, like a flow chart where every step is its own little block. Now a good example of this is a guy I've worked with named Felix. Now he's in the investing nation. He built this really simple system where he would scrape posts off of Twitter and then feed them into AI such as Cloud to help create his content. But here's the key, right? He didn't over-engineer it into some massive, fragile machine. It was simple, it was modular. And when something needed to change, he could just swap that one piece. And that simple, swappable system helped him climb all the way to the top of his niche. And by the way, this lines up perfectly with what McKinsey found, that the companies actually getting value from AI weren't just bolting a tool on top of their old ways of doing things. They redesigned the workflow around it. They built it in modularly like a Lego. So here's how to do this. When you set something up with Cloud, break it into separate steps instead of one giant mega prompt. Save each step on its own, like a save prompt or a project for each piece. And that way, when a better model or a better method drops, you only update that one step inside of rebuilding everything. So build in bricks, not in concrete. Now, by the way, when you're building things in a step-by-step fashion, like with a flow chart, in many cases, you want to build quality control steps into each step of the process, right? So you don't move on to the next step until you know that that particular step of the process is actually good. Because if you don't do it like that, if even one step messes up, then the end product is going to end up being suboptimal. But this brings us to rule number four, which is the lazy 80. Okay, so this is the good old 80-20 rule. And this one's going to feel almost too easy, right? You've probably heard of the 80-20 rule or the Pareto principle, which is 80% of your results come from 20% of your effort. Now, here's how this is applied to AI. If you can get 90% of the result with something simple and fast, do not build the insanely complicated thing just to squeeze out that last little bit. It's kind of like working out, right? You don't need the perfect, scientifically optimized, color-coded program with 47 different exercises to get in shape. Showing up and hitting the basics, squats, sleep, protein, getting some exercise in like playing some pickleball, for instance, gets you 90% of the way there. And the guy doing this simple thing consistently is going to absolutely crush the guy designing the perfect program that he never actually starts. Or he does it, but he doesn't do it consistently. So I've watched people spend weeks building this perfect, elaborate AI automation to save themselves 10 minutes a week. That math just does not work. That is not being smart. That's just expensive procrastination. And the MIT report found that more than half of a company's AI budgets were going into flashy, exciting stuff, like sales and marketing tools. But the biggest actual return was actually in the boring back-office automation that nobody talked about, right? So people were chasing the shiny 20% and completely ignoring the simple 80% that actually paid off. So here's the move. Before you build anything with AI, ask yourself one question. What is the laziest and simplest version of this that's still gonna get me most of the result? Then, build that. Ship it. If you genuinely need the fancy version later, you can add it. But nine times out of 10, you won't need to. So get the 80, then skip the grind for the last five. All right, rule number five, and this might be the real cheat code, is speed. The speed at which you implement is more important than getting it perfect because your first version is never going to be perfect anyways, right? Whatever you build first is gonna be kinda bad. And that's fine. That is normal. The winners don't avoid that. They race straight through it. They ship the ugly version fast, see what's broken, and then fix it live as they go along. And this is true with companies as well. Airbnb started off as a crappy web forum. YouTube started off as a video dating site. And Slack started off as a failed video game, right? It's kinda like learning how to ride a bike. You could read every book in the world about riding bikes. You could watch all the documentaries in the world about how to ride a bike. You could hire whoever won the Tour de France most recently to teach you how to ride a bike. But until you actually jump on the bike and start trying to ride it, you are not going to know how to ride a bike. It doesn't matter if you try to study it for 20 years, right? So getting started as quickly as possible and shipping a minimum viable product is the way to go. And speed is king, okay? Speed is the most important thing. The speed of implementation is pretty much directly proportional to how successful you are going to be in life and with AI as well. So the reps are where the growth is, not the planning, but the doing. So stop polishing. Stop worrying about whether the UI is pretty. Stop worrying about all that other stuff. Just ship it. Now you might be thinking, but what if it's not good enough yet? It's not and it never will be with version one. You make it good by shipping it and fixing it, not by hiding it. And this is exactly how I operate, right? When I use AI for something new, my first version is almost always rough, but I get it out the door fast. I see what works and then I improve it from there, right? Ship a minimum viable product as fast as you can and then stress test it and then tweak it and modify it from there. And remember those startups from MIT that went from zero to 20 million? Well, they did not spend a year polishing in the lab. They shipped, they got real feedback and they iterated. So here's how to start this today. Pick that one task. Give Claude one solid attempt at it and then actually use the results, even if it's only 80% there. And don't sit on it for a week trying to perfect it. Ship it, watch what breaks and fix that. Done beats perfect every single time. Now, like I said before, we've been working on AI for YouTube. We might release it as a product, but right now we're just letting our clients use it. And we think it's the absolute best AI stuff that is available on the market by a long shot. With that being said, here's a few other tips on how to use AI really well. One, let AI interview you. Tell it the outcome that you want. Try to give it an example of the outcome that you want. Then let it interview you and give it as much context or as much information as it needs in order to achieve that outcome. Two, don't become dogmatic about AI tools. Right now, Claude is the best when it comes to content creation, but Codex is actually really good when it comes to coding. Claude code is still really good, but Codex is up there with it. Many people would say that it's better. A few months from now, maybe Claude code is gonna be better again. But even within Claude, there's a bunch of different Claude tools. We have actually found that sometimes Claude code is actually worse than just using normal Claude or Claude projects for certain types of tasks. But we wouldn't know that if we became dogmatic about having to use Claude code for everything. Additionally, there's some things that Claude still isn't the best at. Codex is better right now. And additionally, when it comes to visual stuff, ChatGPT and Gemini are both better at visuals, right? So if you're trying to make a thumbnail, ChatGPT and Gemini are better right now. But the cool thing is you can actually build that into the software that you create, right? Or you can just build it into the workflow that you create. And then the third tip is whatever you're using Claude for to build, right? Whether it's building some sort of software app or using it for content creation or using it to do agency work or whatever you're using it for, document in public. Make videos documenting and showing the cool stuff that you're doing with Claude. There's been a lot of people that have blown up their personal brands just doing this and you don't need to have fancy videos. A lot of the time, it's as simple as just making a Loom video and posting it on YouTube. And I talk about why you should do that and why it's such an asymmetric bet to do that. Because if you build a personal brand, you own the distribution. And over the years, you're gonna be able to make millions of dollars from your personal brand, even if it's a tiny one, whether or not your products win or fail. And it's gonna make it much more likely for you to be able to actually launch a successful product if you have a personal brand to launch it from, right? So this is an asymmetric bet. If you win, you win. And if you lose, you still win because you built up a personal brand. So it's an absolute no-brainer. And of course, the best place to build a personal brand, the mother platform, is YouTube. And that is what I actually teach people to do. I've gotten this channel that you're watching right now to make over $10 million just from YouTube alone. And I've actually worked with clients and helped them generate over $100 million from YouTube. And the typical types of people I work with are business owners, both online, as well as some types of local or localized business owners. But typically, it's the type of businesses that get clients, right? So we're talking about law firms, cosmetic dentistries, realtors, and financial advisors. So yeah, both online businesses, as well as some types of in-person businesses, YouTubers who are getting views, but they're struggling with monetization, YouTubers who are crushing it, but they want to crush it even harder, or people who want to treat YouTube like a business from the beginning, right? They're very serious people. They want to give value to the world. They want to solve problems for the world. They want to make educational content and treat YouTube like a business from the beginning. So if that sounds like you, you can go ahead and apply to work with us by clicking the link in the description and the pinned comment below. We are accepting three to five people right now, but we only accept about 18% of people who apply to work with us. So only do it if you're very serious about growing and making money on YouTube. And if you want to see someone that we helped fully automate their channel, literally fully automate their channel to the point where they don't even have to record themselves anymore, you can check out this interview that I did with my client, Thomas, by clicking right here.