Today I'm going to be building an AI SAS product right in front of you guys pretty much live. I'm going to be basically starting at the complete beginning. I don't even have an idea yet. I'm going to go through ideation, planning, building, testing, authentication, payment processing, connecting it to a real domain. I'm going to do all of that in front of you guys so that after this video you basically know exactly how you can find some sort of idea, turn it into a product, and then actually go look at getting some customers for your product. So, I don't want to waste any time. Let's just get straight into the video. So there's six major P's that we're going to be focusing on today when we're building this sort of like SAS sprint. So the first P is pain. I'm basically going to pretend that I am in your shoes right now. You might be starting from completely zero, a little bit of AI knowledge, but you don't have an idea or a product. So that's what I want to kind of simulate today. So pain, we obviously want to build something that people will actually use. Now in a perfect world, the way that I would actually go about this is I would go get feedback. I would say, "Hey, here's my business idea. How much would you pay for this? Would you pay for this?" and I would get a bunch of validation before I actually build or I'd build like a little P proof of concept to just say, "Hey, here's a demo. When I build this out for real, would you build this? Would you guys find this value?" And then I would go build it. So, basically sell before you build. So, we're going to be using AI here to help us find some pain points and find the business idea. From there, we're going to step into promise. It has to be super clear and basically one sentence on the landing page. What does this tool do? How does this tool promise that it's going to answer the pain that we already found? From there, we're going to be actually doing the product. We're going to be planning it out. We're going to be building it. We're going to be testing it. And then, you know, that's basically the actual core product that we're building today. The fourth one is the plumbing. So, we're going to have to wire up things like the payment processing, things like the authentication, managing the database. Obviously, when you have a product like this, the goal is that you one day can scale it to, you know, potentially thousands and thousands of users. So, you need to make sure the plumbing is there to actually handle that. The fifth one we have is the packaging. So, making sure we have stuff like the brand guidelines, consistent colors, you know, it just feels like a professional product that someone could actually like maybe go to the website and just buy or subscribe to. And then the last one is proof. So, basically all throughout this whole thing, we're going to be doing so much verification because at the end of the day, I don't know how to read Python. I don't know how to code in Python. I don't know what other languages we might use, but AI can do it really well. But it's on us to be the project manager, to be the one that's saying, "Hey, verify this until you're like confident that it's done and prove to me that you're confident." So, it's all about how can we actually utilize the more intelligent and more capable coding agents that we're going to be able to use, but we still remain in control of the judgment. And you know, at the end of the day, if the product fails, it's on us. We can't blame AI, we blame ourselves. So, I'm sure you guys are curious about what tech stack we're going to be using today. We're going to be using three main ones to sort of drive everything, but I'm sure there's going to be, you know, other APIs or other payment processors. There's going to be a lot of things that go into this build, but the three core that I want to call to your attention are we're going to be using codec. So, I have the chat GBT desktop app that I'm going to be using. I'm going to be using codec through this. I'm also going to be using Claude, so Claude code inside of the Claude desktop app. So, you can download that as well. And then finally, we're going to be using Glido to actually talk to everything. It's basically going to be the voice layer so that I'm not going to be bottlenecked by typing. I'm just going to be able to talk as fast as I can. and do this as fast as I can. So, if you want to sort of follow along with this video, then I would definitely recommend you get all three of these things so that you can basically move quick and you can have different AI models and different coding harnesses get different perspectives on your build, which I think is really helpful. If you prefer codecs and you just want to use codecs or if you prefer claw and you just want to use claw, that's totally fine. But I found it to be very helpful to be able to have different coding harnesses and models take a peek at what I'm doing and help me sort of like play devil's 

Right now, most businesses are running on stuff that only exists inside of one person's head. And usually that person is the owner or perhaps you. So, every real decision, it waits on that specific person. And if they were ever going to leave, that information leaves with them as well. And this is not just something you could buy or hire a way out of because it was never a software or hiring problem. So, I came up with a solution. Every price, every process, every judgment call, it's inside of one folder the company actually owns. And consequently, that whole team is going to be working out of that one folder. So now it's just one brain shared by everybody in a company. And unlike your best employees, this one does not resign. It does not take weekends off. It doesn't walk out of the door with 15 years of customer knowledge. So I asked it, "What happens if the operation manager quits tomorrow?" If you're not familiar with our channel, my name is Nick. I've deployed AI inside of more than 60 businesses over the last 2 years, and this is what owners have been asking me for the most. And it's not just simply enterprise software. It's actually just claude skills and a folder. You just need one afternoon. It requires no coding. and I'm going to be building it all in front of you. And the step that actually gets this working in your organization, getting it across the finish line, it's just having a conversation with the AI system. I'll be showing that today. Let's get into it. And fundamentally, every business that I have walked into has had the same disease, and it actually shows up in four different ways. Problem number one, the owner is the database. This is where you're the only person who can price the big order, approve that exception, answer the hard questions, whatever it may be. which means that every one of those questions it's going to be waiting for you. There's so many things in approval loop that is contingent on one person. I recently just talked to a founder running a sevenfigure business and he told me that his entire bottleneck was just getting his brain onto paper. Now, he is not the exception in this case. He is every owner that I have practically ever met. Problem number two, the company's memory lives inside of the chat threads. Now, here's a real example of what that actually costs. So, a customer, they ask for volume pricing Saturday morning. Now they're expanding. This is good news for them. This is money. However, the question it just sits there all weekend because the one person who knows the answer, they're away. They cannot answer them right now. So by Monday afternoon, a competitor, they've already sent them that quote. And that's not a software problem. That's a knowledge trapped in one person's head. And just because of that, they weren't able to close that deal. Problem number three, nothing is compounding. So every new hire starts from zero and takes months to actually be useful because the training material it's just stuck inside of other people's memories. So when somebody good leaves their 15 years of experience it's leaving with them and just leaving the company in a hole. And problem number four nobody knows what actually matters today. Everyone's busy but the priority is they live in the same place that everything else lives which is somebody's head. So the team they work on what feels urgent instead of what actually is the pressing matters. Now, here's the diagnosis because these four problems, they're actually just one problem. None of this is a software issue. You could buy more software tomorrow if you would like and all four they're going to be surviving because the real issue, it's actually the knowledge. Who holds that? Who owns it? So software, it's basically software right now, your company's knowledge, it is not accounted for. And this is why your standard fix or solutions to this actually fails. The standard fix is typically renting more software. So maybe you implement a new CRM or maybe you get a help desk inside of another department or maybe a task tool or a chatbot tool and now you're past $2,000 a month forever. And the knowledge that you feed those tools, it lives in their database, not yours. So it's not proprietary. Now, what I'm about to be building and showing you instead lives in a folder that you are owning. It's going to be proprietary. But before I build anything, let me just take a step back and show you the entire system. So when we get into the walkthrough, you know exactly what is going on. Now your company's brain, it is effectively a folder. That's it. That is the big secret with this. So the files inside of this, these are what the company actually knows. So this is going to comprise of your prices, your customers, your processes, and even your judgment calls. Now the skills, these are how the company actually does things. So the rules for using that knowledge. And every application, Claude included, it is just a window you look at that fol

Anthropic, the company that created Claude, recently sorted through hundreds of their own Claude skills. And the biggest quality jump didn't come from the skills that do the work. It came from the skills that check it. If you're new here, I'm Dylan. I run an AI consultancy. And when I sit down with my coaching clients and we go through what they've built, there's almost never a checking skill in the list. Nearly everyone builds the same two kinds of skills and stop. So, what I'm going to do is I'm going to walk you through all four skill types and show you which ones you're missing and I'll give you the prompts to fix it this week. So, let's get into it. So, they, Anthropic, reviewed hundreds of skills inside the company. After they reviewed it, they then published a blog that covered the nine types of skills specifically targeting engineers and developers. So, after I read that blog, I then distilled this down into four skills that I feel are suitable for non-technical users. And it also highlights the two skills that Anthropic themselves find the most valuable. Now, first, what are the two skills most people are already using? So, the first one, this is probably the only skill most people build, are doer skills. So, a doer skill is simply a skill that automates a given task. So, there's probably something that you do on a reoccurring basis that if you encapsulate that into a skill, it can then do that for you automatically. Some simple examples are if you have a reoccurring meeting on Mondays, it can write the follow-up email for you perfectly. Or maybe when there's a specific service or product that you sell, the AI can draft the proposals for you for new clients. That's the most common type of skill that I see and most people build. The second category of skill are formatters. So, this is the second most frequent thing that I see that people build with skills. And this is when the AI formats a document to match your specific brand in a pixel-perfect way. And the reason this is ideal for skills is because you can have AI embed code into a skill, which increases the chances that it matches your format for a presentation or report perfectly. This is probably a good thing for you to remember as well is that if you want AI to match your brand guidelines for a deck or report, go ahead and use a skill for that. Now, these are the two things most people build. So, when I look at somebody's library of skills, when I look at their shelves in their library of skills, I see most often that doers are the most often type of skill, and the formatters are the second category. But, very rarely, if ever, do I see the other two categories. That's what I want to focus on today in this video. Quick pause in your regular programming. This video is brought to you by me, as always. Two quick things. First off, below is a 30-day AI Insight series, completely free. You'll get 30 insights in your inbox of how you can apply AI to your business and your work. The second thing is if you'd like to work with me, below are a series of offerings to see if there's a good fit between the two of us. Now, let's get back in the video. So, the first category of skill, and probably the most important skill type that Anthropic labeled inside their blog, are checker skills. So, the purpose of a checker skill is simply to check the output of another AI. And for most of the people I work with, the highest value version of this skill is simply checking the data an AI extracted out of a document. Cuz often times, we have AI extracting information from PDFs, from PowerPoints, Excel sheets, and then putting it somewhere else. We always have to double-check and audit the output to make sure it did it correctly. Well, if you have a second skill that's a checker skill that's checking that AI's output, the likelihood of that being correct is much higher. And that's why this is so valuable and so important. Now, both with this skill and the next skill, I'm going to give a bit of a caveat. Before you actually separate these out as separate skills, I'd recommend adding them as simple instructions into your existing doer skills. That's where this all should start. But, later on, I'll tell you when to actually evolve that instruction into its own skill. But, note that in the beginning, it's best suited as just an additional line inside of your instructions for any skill. So, how do we actually create this additional line or lines inside of our instructions for a checker skill? Well, here's a prompt you can steal. All this prompt is doing is simply giving an AI a good output. So, here we're saying, "I want you to study this approved proposal." Now, this use case is dedicated to proposals, but the output could be anything. You just want the AI to see what good looks like at the end. After it looks at it, you want it to methodically understand specifically what makes this good, and then extract out from that a a checklist of things it can check on future instance

So thanks to AI, it is now possible to build a one-person business, but only if you know how to use Cloud Code the right way. So in this video, we're going to cover what this business actually is and how to build it step-by-step. So before we get into anything, let me just tell you about what this business is because there is a lot of noise out there right now and there's so many different opportunities and most of it just makes this whole thing sound way more complicated than it really is. So the business model I'm talking about is becoming an AI consultant. It just takes one person, no agency, and no team. Just you sitting at your computer using Cloud Code to help businesses use AI inside their actual operations. And I know the term AI consultant might sound a little bit vague, so let me be more specific. You're not saying you're an AI builder. You're not saying you're the automation guy. The framing that actually wins in 2026 is being an AI partner because builders sell features, but partners sell real outcomes and businesses obviously pay way more for outcomes than they do for fancy features. So the identity that you really want to take on here, which is the AI partner or AI consultant, it really does matter a lot. Now real quick, if you don't know who I am, my name is Nate. I scaled my AI agency to over $100,000 a month and then I exited that business and now I run a free community of over 375 people building with AI. I've worked with real estate agencies, HVAC companies, coaches, marketing agencies, a ton of different businesses, and I've seen the same patterns over and over. So when I tell you guys this is doable, it really is and I've seen it work for me and for hundreds of my students. And the reason Cloud Code unlocks this in 2026, specifically, is very simple. It's just natural language being the interface, so you don't have to be a developer. You don't need a formal computer science degree. You don't need to know how to write code. You just need to be able to describe what you want clearly and obviously find the opportunities for automation clearly. And the build time has just collapsed, you know, like stuff that used to take me 2 hours in the end to build. Now it takes me about 20 to 30 minutes to build in Cloud Code, sometimes even less. So the point is the barrier to entry is dropping and it keeps dropping and the leverage that you can get from one person has gone way up and it's going to continue to go up. So you might be wondering, okay, cool, but what am I specifically doing as an AI consultant for these different businesses? Like what does that actual work look like? And that is what we're about to get into next. Okay, so every single business in the world is trying to move one of these three buckets. So, number one is to get more customers. New leads, more booked appointments, more conversions from a specific source, anything that brings somebody new into that business's ecosystem who wasn't there before. And this first bucket is the most growth-focused. You know, businesses spend the majority of their time and attention and money towards this bucket. That's where their head is most of the time. Now, number two is to make each customer worth more. So, the average order value, the lifetime value, the retention rate, time between repeat purchases, the upsell rate. This is basically the idea of getting customer to pay you more and stay longer. In mature businesses tend to find their highest leverage automations in this bucket because the math on customers that you already have is usually way better than the cost of going out to acquire new ones. And then bucket number three, cut costs. Hours per task, error rate, ticket count, time to completion. Essentially, you want the same outcome, but you want that outcome to happen with less labor, less rework, less waste, less money. And this is where most operations-focused automations land because the metrics are easy to baseline and the wins are easy to attribute. Like if you save somebody 20 hours a week, you can prove that on a calendar. So, the idea here is that every project that you go for should roll up into one of those three buckets. If a build doesn't clearly move the needle in one of those buckets, then it's probably not a project that you want to pursue because it's going to be way harder for you to prove the value that you have added to that business. So, as you start thinking about what do businesses actually want and what do they actually pay for, you always want to route it back to one of those three buckets. So, just to put some concrete examples on it. Lead qualification systems, automated follow-up, that is get more customers. CRM automation, onboarding flows, that's make each customer worth more. Internal knowledge assistants, reporting dashboards, ticket sorting, that's cut costs. Now, with something like onboarding, for example, you could argue that that is making each customer worth more, but also cutting costs. So, sometimes

Chat GBT just changed how we do work forever. It's called record and replay where you just hit record, you go through any task on your computer one time, and it turns what you just did into a skill that it can run for you from then on. You don't even need to prompt or instruct anything and you just do this once. OpenI's own examples are filing an expense report and even booking time off. But think bigger than that. So the weekly report somebody on your team spends two hours polling or even the onboarding checklist that only one person knows how to do. You just record it once and it belongs to the software. So which one do you record first? Well, it has to happen every single week and it has to take more than 20 minutes. And it has to be the same steps every single time. So check all three of those off or to skip it because anything that changes every time still needs a person. But do this with your three worst weekly jobs and you will get a full day back every single week without hiring anybody new. Come at need and I'll send you all the resources to get started.