Give me 16 Minutes And I’ll Teach You How To Pick an AI Business That Never Fails

Most AI businesses started in the last 18 months will likely be dead by 2028. And that is not intended to bring any fear. It's just a pattern. We are standing at the same point in the AI cycle that the internet was in March 2000. A lot of hype happened. The wash out started. By 2004, half of the companies that existed in 2000 were gone. And the ones that survived didn't have better ideas. They just had a different structure if you look at them. And what I did was I spent the last two years watching founders pour their savings into AI businesses that look promising. And I can usually tell you in about 60 seconds which ones won't exist in 3 years. I'm not clairvoyant. It's just because the survivors share a structure and the casualties share a different one. So in the next 14 minutes or so, I will give you the three filters that separate them. They come from Stanford, from Netflix's boardroom, and one of the most underrated economists alive. Almost nobody on this platform is using them, but together they form what I call the durability triangle. So, if your idea fails any one of these three filters, it will likely not survive the next cycle. The good news is by the end of this video, you will be able to run any AI business idea through all three filters in under 90 seconds. So stick around for filter 3 because it's the one that nobody's talking about and it's the reason why your grandmother's accountant still has a job. But let's start with something important. There is a lie that almost everyone online is trying to sell us. Pick the right niche and you win. But that's purely wrong. Niche is downstream of structure. Actually Shikar's Gosh's um Harvard Business School research found that 75% of venture-backed startups fail to return capital. I mean think about it. These are companies with professional investors vetting capital and what looked like a good niche to begin with. The failure rate is not because the niche was wrong. It's essentially because the underlying structure of the business itself could not hold weight when conditions changed. And for AI businesses, conditions change every other week. There's a new model release, new pricing tier, new competitor shipping the same rapper. The niche doesn't save you. But structure can. And here's the thing. structure is basically invisible until it's tested. You can't see whether a business has durability the way that you can see whether it has a logo or a website, right? You have to know what to look for. And almost nobody's teaching that part. They're teaching what to build, but I want to teach you what to refuse basically. And like I said, there are three filters that I've put in a framework that I call the durability triangle. Filter number one is the phase filter. This one is built on Carla Paris's research on technological revolutions. This one basically tells you whether you're building for the hype phase or for the workflow phase. Then we've got the power filter built on Hamilton Helmer's Seven Powers, the strategy book that Reed Hastings put in front of Netflix's board. And this one tells you whether you have a structural advantage or just a feature. And then the third one, I call it the permanence filter. And this one is built on Nasim Taleb's Lindy effect. This one tells you whether the problem that you're solving will still exist long enough to make solving it worth your time and your effort and your investment. Now, each of these filters cuts the field by roughly 80%. Combined, you will essentially be left with the ideas that actually compound. So, let me walk you through each one in a lot more detail so you actually know how to apply them. So, filter number one, like I said, is the phase filter. Um, and I feel like I need to give you a very, very quick background. In 2002, a Venezuelan economist called Carlo Perez uh published a book that was called or is called technological revolutions and financial capital. It is required reading at the London School of Economics at Cambridge and actually inside the UK government strategy unit. And almost nobody outside academia has heard of it. But basically her finding was that every major technology whether it was steam, electricity, the car, the internet goes through the same two phases. the installation phase where capital floods in, valuations explode and hype peaks and also most startups die. And the deployment phase, that one is slower and less glamorous, but it's where the actual durable businesses get built. And if we look back at the 2000.com crash, that wasn't the death of the internet. It was the transition between phases, right? Amazon, Google, eBay, all of them survived because they were building for deployment, not for installation. The companies that did die though were building rappers on the hype itself. We are at the exact same transition with AI right now. Okay, so here's the test. You look at your idea and ask one question. Am I building infrastructure for someone's daily work or am I building a thing that serves the hype? A chat rapper that summarizes documents, it's hype. An agent that lives inside a law firm's workflow takes the same actions that a junior associate takes and would be harder to remove than a parallegal, that's deployment. Let's think a prompt marketplace for example that's clearly hype but a vertical SAS for HVAC contractors that handles the quoting the scheduling the customer follow-up clearly deployment. The difference is not the sophistication or the number of lines that your code has or how sexy your landing page is. The difference is whether the business survives when the model underneath it changes. Deployment businesses are independent of which model wins. installation businesses die the exact day that the model goes to be replaced. And I want to give you an example. A few months ago, we ran an ideation workshop with a founder who had built a really sharp AI tool. It was beautiful execution. 3 weeks later, a major model release made his entire product redundant overnight. Sounds like the end, right? And it feels really terrible to be in their shoes, but that's because they had built on installation. Okay. So what you would need to do and what we did was to rebuild the project to live within their client's existing workflow software. So basically provide the same skill but a different placement. And now the model can change as many times as it wants and the business will still exist. So now you can have the real conversation about niche and pricing and distribution but you're starting from a structure that will hold weight. And if you're serious about actually building one of these durable AI businesses, not just running ideas through the triangle and then staring at the blank screen wondering where do you start, then this free course from HubSpot is going to save you weeks of figuring out how to build your first AI agent, for example. It is called Build Your First AI Agent, and it has 18 hands-on videos that walk you through building real working AI agents from scratch, built for the exact moment that this video sets up. You found the idea that passes the triangle. You've got a durable business worth building and now you actually ship the first version. Here's the thing that most people get wrong once they finally picked a durable idea. They spend weeks, if not months, deciding which platform to use and never write a single prompt and then end up with nothing but tabs. You don't have to because HubSpot already wrote the walkthroughs. You can copy the prompts, follow the build steps, and ship the agent. Like I said, inside you're going to find 18 walkthroughs that map to what most of you actually need. The first one builds a personal assistant agent in NA10, which is drag and drop. There's no coding connected to your real calendar and your inbox. There's a content repurposing agent built in make that turns one YouTube script into LinkedIn posts and short form video scripts automatically. And my favorite is a customer support bot that is trained on your entire website's knowledge base or your clients deployed in about 12 minutes using Chatbase. If you watched this video and thought, great, I've got a great idea and it passes the triangle, but what do I actually build first? This is your next step. Hit the link below to grab build your first AI agent. It is absolutely completely 100% free. And big thanks to HubSpot, by the way, for partnering with us on this video and for making this resource available to everyone. Now, filter number two. Like I said, this is the power filter. And again, I feel like I need to give you the context. In 2016, a man called Hamilton Helmer published a book that was called Seven Powers. He actually had advised Netflix for decades. And in fact, Reed Hastings calls it the most important strategy book that he had ever read. It is in fact taught at Stanford. Uh, and Stripe's founders also cite it and Spotify's board uses it as a frame of reference too. And basically, if I were to summarize that and give you uh Helmer's argument in a couple sentences, basically he says that a business is only durable if it has at least one of seven structural advantages. Again, structural. We're not talking about features or vibes or, I don't know, beautiful websites. We're talking about the kind of structure that a competitor cannot replicate by hiring smarter people or shipping faster. There are, like I said, seven in the book, but for AI businesses at the size that most people are building, only three are realistic. The first is switching costs. The longer a customer uses your product, the harder it becomes to leave. Their data lives there, their workflow is wired into it, their team is trained on it. The minute that you build something that a customer would have to rip out and reconfigure to replace, you've got a real competitive advantage. The second one is process power. Your way of doing the job is faster and harder to copy than your software. Toyota, for example, has a process power. McDonald's has process power. And basically, if you look at them, the pattern is the same. It took years to develop. Anyone can see what they do, but almost nobody can replicate it. And for AI, process power looks like a tested workflow that gets a measurable result that your competitors cannot match. Not because they don't have AI or they don't see the workflow. It's just they don't have the recipe. And the third one is a cornered resource. Basically, it means that you have something that no one else can access. Whether that's proprietary data or exclusive distribution or maybe a founder that the market already trusts by name or you have a big audience that people lean in to listen when you say something, that is a cornered resource. If you own the only label data set of a specific industry's workflows, that's a cornered resource. Tesla, like many people know, is not a car manufacturer, is a data company, and that's their cornered resource. So, if you want to test, you need to think and name your power out loud in one sentence because if you can't, you don't have a business yet. You have a great feature at this point. Now, filter number three is the permanence filter. This one is built, like I said, on Nasim Taleb's Lindy effect. And the Lindy effect basically says the longer a thing has already existed, the longer it will continue to exist, probably. So essentially things with long histories are more likely to survive than things with short ones. It's why classic literature outlasts bestsellers. And if you think about applying that to business, the longer a problem has existed, the longer it will continue to exist, probably. So you want to build solutions for problems older than 1990. Bookkeeping, hiring, lease abstraction, customer follow-up, insurance claims, inventory management, sales pipeline reconciliation. These were problems in 1985 and they will continue to be problems probably in 2045. The AI just changes. Who solves them faster? So the test is was this a problem before the internet existed? If yes, it passes the test. If no, you probably need to rethink it. This is the filter that should scare people the most because it's the one that exposes how much of the AI business advice on this platform is built on problems that won't exist for long. So, let me push three popular AI business ideas through the durability triangle. So, you can see the filter actually at work. Uh, let's take idea one, an AI agent that writes LinkedIn posts for executives. Okay. The phase filter, workflow embedding sits inside their daily output. Pass. The power filter switching cost are real if the agent learns the executives's voice over time. Pass. Permanence filter. Was thought content for executives a problem before the internet? Marginal. Ghost writers wrote books and op-eds for executives in 1985. So somehow the underlying job or problem is old, but I would pass it with caution. This idea survives the triangle. The challenge becomes execution, not durability. Let's take another idea. A general AI chatbot for small businesses to handle email. Phase filter. This is a feature, not infrastructure. Every email client will ship this natively within 18 months. So I'd say it's a fail. Just move on. Idea number three, let's think about um AI powered platform for HVAC contractors that handles quoting and scheduling and customer follow-up build monthly. Let's say phase filter. This one sits inside their daily operations. So, it's a pass. Uh power filter switching costs are enormous once the contractor's customer database lives inside it. So, definitely pass. And then last, permanence filter. Clearly, HVAC contractors needed quoting and scheduling in 1985 and they will probably need it in 2045. So, it's a pass. That idea is structurally durable. Now, you can have the real conversation moving forward about niche and pricing and distribution, but you're starting from a structure that holds weight, like I said. So, that's the framework. Now, here's how this changes what you decide to build. The durability triangle does not tell you what to make. If you want ideas, make sure you subscribe because I have a few videos coming up where we talk exactly about that. But what this one does is it allows you to see that refusal is the most underused leverage in business. Every no that you give to a fragile idea is a yes that you'll be able to give to a durable one with full conviction and all your time and focus and effort. Because when you build with this filter in mind, your business outcomes compound differently. Retention is higher because switching costs are real. Pricing power is higher because the alternative is harder to construct and build. And acquisition is cheaper because you're solving a permanent problem that people already search for solutions for. So you're not building faster. You are building once. Now obviously like I always say, results depend on your implementation and your niche and the work that you put in. But the gap between people running this filter and people not running it shows up around probably month 10 to 12. The ones running it still have a business. The ones who weren't running it, they're probably rebuilding from scratch. Now, if you want the onepage durability triangle checklist, the same version that we use and run with client ideas or new projects that we come up with, you will find everything in the community. The link is going to be here as well as in the description down below. We have two communities actually. One is the free one, the Trailblazers Hive. There you can come and be joined by thousands and thousands of other people who are on the same path as you and who are looking to learn and apply AI specifically in their business or start a business with AI. So if you're at that point, make sure you come and join us there. It's completely free. We have calls every two weeks. We have free challenges that you will get a ton of value out of. I can promise you that. But if you want to be handheld and you want a more specific step-by-step approach on how to start or or grow your business with AI, then we have the Founders Hive. And that one not only has our 90-day step-by-step process, but it also has an upcoming Claude course that we have been building and we are releasing probably at the time you're watching this video. And you'll be able to use Claude to do all of this for you in your business. So, if that sounds interesting, like I said, the link is here as well as in the description. And until next time, thank you so so much for watching. Like this video if you did. Be sure to subscribe if you haven't done so. Share it with anyone in your circle of friends or family or co-workers who is thinking about starting an AI business, but is not really sure how to move forward and they fear the risk of failure. I think this video is perfect for them. Thank you again, and until next time, make sure you go ahead and watch this video here. I'll see you there. Bye.