Why 90% of AI Startups Are Solving the Wrong Problem

Are you getting a bit overwhelmed by all the AI agent talk? Look, I've been in tech for 40-some-odd years. I was on IBM's secret AIX team when Unix was literally classified. And what I'm about to show you, this changes everything you think you know about AI success. Because the real AI revolution, it's not happening where you think it is. I'm Dennis Lewis from ratio.dev, and here's something most people don't realize about AI right now. Everyone's Talking about ChatGPT, Claude, Gemini, the models and the layer, the first layer on top of those models, which are the agents now that are able to orchestrate those models in a more powerful way. But that's still like focusing on the engine when you're ignoring the whole car that's actually going to take you someplace. The truth is AI without the application level is just pretty much expensive plumbing. And today I'm going to show you why 90% of AI startups are solving the wrong problem. I'm going to talk to you about the 3-layer stack that separates the winners from the losers and the counterintuitive reason why the dumbest AI applications are making the most money. But first, let me tell you about a massive failure that taught me this lesson the hard way. So back in the 1990s, I moved to Spain fresh from working at IBM in the laboratory in Austin, Texas. And hey, I thought I was a really hot commodity. You know, I had all this technical knowledge that was really pretty hard to come by. And so I get to Spain, I get hired by a local IBM business partner, and I start watching the Spanish entrepreneurs, the guys who barely understood email, and they were crushing it in the tech business. And meanwhile, there I am in the tech department with all my amazing Unix experience, my knowledge of operating systems and programming languages, stuff that people would kill for, and nobody's really paying that much attention. And you know what the difference was? They weren't selling technology. They were solving real problems. One guy I knew, Miguel, he built a super simple database for restaurants to manage their reservations. It was nothing fancy, believe me, but he understood restaurant owners and he spoke their language. And I realized there that I was trying to sell what I knew and he was selling what they needed. And that is a really important lesson that I learned in those early days. And that's exactly what's happening in AI right now, except the gap is even bigger. So all of these big companies are teaching you about what they know, what their model does, what their killer feature is, and nobody's really focusing on solving the problems. So most successful AI companies really operate on a 3-layer stack. So on layer 1, that's where you find these LLMs. You find ChatGPT, Claude, Perplexity, Gemini, the, the new agent orchestration level that's kind of glued to them now. And this is what everybody's obsessing over. And I get it. It's cool. I mean, I am a tech geeky nerd, kind of, uh, you know, up there with, with all the rest. And I, and I can get enthusiastic about this sometimes as well, but you know, this is where they're spending literally hundreds of billions of dollars, you know, engineers every day, there's some new thing coming out. Now we've got Cloud Mythos, which is, you know, taking down servers all over the, all over the internet. But the truth is this bottom layer is really rapidly becoming a commodity. It's, it's ending up to being like electricity. It's enormously powerful. It's really cool, but it really doesn't differentiate anymore. And, and I'm a little bit tongue-in-cheek there because obviously there are some really cool innovations that are constantly coming out. Layer 2, what I call the integration layer is where it really starts getting interesting. And this is where AI actually meets your business systems. Okay. And by that, I mean, sure. Right now, you know, and this is where most AI startups are kind of focused right now. They're, they're working to get the fluffy part of your business integrated with AI. So, you know, we're talking your marketing collaterals, your landing pages, the, your brand voice. The content creation that you're creating for social media or for YouTube. And that is a really, you know, that's a, that's a great place to be. It's, I'm not saying that's not super effective. I mean, we use it every day, but where it really starts getting more powerful is when you start bridging the gap with the actual transactional data that is making your business run. Okay. And every business has their own systems. So if you're a mom-and-pop pizza shop, you've got a point of sale system. You've got somebody, a bookkeeper that's keeping track of your accounting. You've got, you know, probably some sort of a scheduling system for the delivery people and the people that are in the store. You've got HR data. It's, you know, and that's on the smallest scale. If you go up a bit and you're in a more middle-sized sort of a small medium-sized business, you've probably got an ERP, you've got invoicing, you've got CRM systems, you've got project management systems. And inside this structured data that runs your business, it's like it's siloed away from all this wave of stuff that's happening on the internet called AI. And the real power is bridging that gap. And that is the secret AI story that nobody's telling right now that can absolutely explode your business's efficiency and give you a real competitive advantage in the marketplace right now. But most people aren't focused there yet. They haven't thought about this. And that leads us to level 3, which is the application layer. So when you can actually merge and build applications that combine AI, structured data from your ERP management systems and also combine human beings in your business so that the applications they're using are being augmented with all of the AI capabilities that these new models bring to the table. That is absolutely a winning proposition. It is crucial and it will be much more crucial every day. So I want to talk to you about an example. Because I don't want this to be completely theoretical. We have been working on a project management system recently for a customer of ours. And this is tasks, daily reports, subtasks, you know, milestones, due dates, when it was completed. And what we have done is we have woven AI into the guts of this application. And how does it work? Well, first of all, we created an AI persona. Her name is Maggie, and she's able to read and understand everything that's going on in this project management system. So she sees every time a task is postponed. She sees every time a task is completed. Every time somebody leaves a voice note, you know, regarding a project, she parses it and understands it. And then she provides insight to the entire team that's using the application. So that means that not only is she able to be a virtual super project manager, but she sends daily emails. Soon she's going to be actually, you know, sending videos, talking to the employees, telling them, you know, hey, I noticed this on this project. You probably should look at that. But she's filling the gaps. She's, she's able to do so much more because she has access to everything that's going on. So I'm going to jump in and show you. First, I'm going to go into Relevance AI. This is the middleware platform that we use for building out AI functionalities. And here we have a tool that analyzes a project and it takes in a report date, a, the, the ID for the project, a start date and an end date. And what it's going to do, the first thing it's going to do is it's going to pull in all of the project data from the Bubble application that we've built for project management. And then it's going to generate a customer-facing report. It's going to generate an internal report. It's going to calculate a health score for the project. And it's going to return this back, all of this information back to the Bubble application. And we can pick and choose which models we want to use for each step because we have this robust platform in the middle. And you know, you can come in and read the prompts and tweak the prompts so that to make things work better. And then this information is being sent back in from the Bubble application. So it's constantly updated. So every time this AI says, I need to analyze project data, it will pull it from Bubble. It'll analyze it and it'll send back its conclusions to the Bubble application. And the way we do this is by generating JSON structures, which AI models are really good at understanding. To explain it to you, it's a kind of a structured way of expressing data. So it reads from the actual database and presents it in a friendly format for the, for the— here it's bringing the project's name, its description, so the AI knows what the goals are, the ultimate goals of the project are. Then it's, you know, turning back all of the completed milestones, the pending milestones, the completed tasks, the pending tasks, and all of the verbal project updates that have been associated to this project. Soon it will also be returning back the chat conversations that are being handled inside the app as well. So the Slack information that from the daily conversations people are having about the project. And it uses all of that information to create those insights and send them back to, to the Bubble application, which can then take them, generate an automatic PDF format with a really cool report, a cover page, you know, well-formatted and automatically send it by email to the customer, the customer-facing report. And there's, there are some checks in the middle so that the user can tweak the content before it goes to the customer. The internal reports get sent to the internal project managers and then it sends them by email, crafting, you know, clever emails. And all of this happens automatically. Now, this You know, it's like bringing in a world-class, you know, project director who leads all of the project managers, who has instant visibility into everything that's going on in the projects and can provide insights. So about this, this type of task is getting postponed too often, or, you know, you've got bottlenecks here. It can also analyze each employee's individual activity across all projects. You know, if you want to have a weekly or monthly personal review with every employee, it will, it does the same thing and it scores how effective they're being and, you know, where they're falling short and where they're really exceeding expectations. All automized, all happening because the AI has access to this core key information. And it's working on all three of those levels. So we're using the AI models, we're using the agent technology, but on top, we're integrating with critical business information systems. And we're melding it into an application that combines the human employees in the company with the virtual employees that have access to everything, all of the data that's going on. And believe me, it is a git, total game changer. It is so powerful. It's a true competitive advantage. And I think that if we were to take a step back and say, okay, yes, we, you know, it's really cool that yesterday Perplexity announced that they can connect to your, to all your bank accounts. Okay. That's great. You know, and we can pull that information and we can pull in information from QuickBooks. We can pull in information from your project management system. We can pull in information from your ERP and we can let AI sift through all of this information that's coming from different places and give you real-time recommendations on the decisions and the actions that you should be taking to be more effective in your business. If this doesn't sound like a competitive advantage, I don't know, you know, another, what another one is, but nobody is talking about this yet. And I think this is the most powerful AI story that nobody's telling right now. So with that in mind, Bringing to the table no-code tools like Bubble, like Relevance, and then combining them with the most sophisticated AI models is not only powerful, but it's doable. Okay. This is where I think people are missing out that this doesn't require million-dollar investments to make it happen. This can happen with inside very reasonable budgets for your business. It can be fast. We're not talking about, you know, needing months and months of time to make this happen. And it is something that, you know, if you don't do it now, your competitors are going to be doing it. With that in mind, I want to, you know, just kind of close out here talking about, you know, this opens the door to doing things, starting small, finding those core problems that are really holding you back, figuring out how to combine those data that, that, that those core problems need and solving them in your business. So maybe it's something, if you're a real estate company, it's something that is able to pull in all of the information about the listings that you have and map them out and write awesome property descriptions that pull in, you know, the images that your, your agents have taken in the field and create spectacular landing pages. If you are a restaurant company, maybe that it's analyzing every ticket that goes into your, you know, that that you make to figure out what changes you should make in your menu so that you're, you're really focused on what your actual customers are consuming. It doesn't have to be everything all at once. It can be, you know, small steps taken that make big differences. So if this is kind of something that interests you, I totally recommend that you go to ratio.dev and you reach out to us. We'd be more than happy to have a conversation with you about how to employ this in your business, sign up for our free no-code resource center. It's just name and email. You'll get access to vetted plugins, to videos, to podcast episodes, to, you know, to information about my, my books on Amazon, all of that totally for free. Sign up for our school community, Bubble Gods, which is where founders and business owners, just like you, are coming together. We're building a community of people that really are leveraging these technologies in real businesses, in not Silicon Valley startup sort of ways, down to earth, street-level businesses that are harnessing these technologies today to make a difference. And we'd be glad to help you on your journey. So if you've liked this video, give it a thumbs up, subscribe to our channel. I mean, we, that makes us really happy and I will see you on the next video.