Transcription
If I was starting from scratch today, knowing everything I know after growing several six and now a seven-figure business, I would do things really differently. And in this video, I'm going to share with you exactly how I would build a one-person million-dollar a year business and give you the entire playbook and road map I would use to just implement this.
And the truth and maybe surprising bit is this wouldn't be SaaS. It wouldn't be tech-heavy even at all. Certainly wouldn't be a course or even a YouTube channel. It would be finding a very expensive problem that is very painful for real businesses who would pay a lot of money to have it solved. And typically this revolves around making them more money or saving them a bunch of time.
So there's going to be a free starter repo link in the description below. You can check out and just like install this and you're 90% of the way there. But now let's jump into the most important part which isn't how to do the tech or how to do the business but where to start because I think a lot of us get started down the wrong path and we're kind of destined to fail before we even get started.
And what I'm talking about here is doing things in the right order. You know, a lot of us are techies these days. We want to jump into building an AI agent and this complex workflow and all these automations. The fact is that's prematurely optimizing for a problem you don't yet have because the most important thing you can do is solve a really important problem for people who are willing to pay you money for it. And until you figure that out, you really don't have a business. You just have a project and really a distraction.
And so what you want to do is solve a really expensive problem. And again, you're going to either save people time or make them money. I like making money because it's easier to quantify the return of investment that they'll make on me. I want to find and prove the workflow with one client first. And truth be told, even in the AI era, I would do this as manually as possible. Then you start to optimize it from there. Productize the deliverable. Create agents around this. Really systematize and automate as much of this as you can, but don't start there. Right? So if you build the agent first, you're building before you really even know what the problem is and who you're serving and how to solve this problem for them. Uh so once you've productized deliverable, then you can build agents for this and stay locked in, right? Don't get distracted by the shiny object and go do the next crazy thing. Uh stay doing exactly what you're doing and provide as much value to those core clients as you can.
And here's the math on a million-dollar a year one-person business. And just as a caveat, I've been in business uh as an online entrepreneur for about 10 years. Never before has this been possible. Never before has there been a better time for you to take exactly what I'm saying today. Just put your blinders on and grind for a few months and you'll be well on your way to this. And what the math looks like is to do a million a year, it's $83,333 a month. If you have retainer clients at $5,000 a month, which for a B2B services business-to-business where you're a business solving real problems for other real businesses, charging five grand a month is nothing. We have clients at my business that Castos that we charge a lot more than that and they're very happy to pay us because even in the podcasting world, we solve a pretty painful problem for them. At $5,000 a month, you need 17 clients. You don't need 17,000 clients. You need 17 clients. And you can manage 17 clients, especially with AI and automations and agents, pretty reasonably. I won't say easy and sugarcoat this for you because to do this will be life-changing, but it's also really hard work. And so, you got to lock in. I'm just giving you the road map to go from it's possible to here's the 80/20 of how you do this. And so, we're going to get started building this here in just a minute. I just want to share with you the stack that we're going to build so you can start to conceptualize how this all fits together.
So, you need tech, right? And all of this is going to live online. So some of it will live locally on your computer but some of it we will deploy to the cloud and that becomes really important as we start to build this. We are going to use the Hermes agent. You want to use OpenClaw, that's fine. If you want to use CodeX, that's fine. I'm using Hermes because for the always-on AI agent this is my tool of choice as of today. This is May 2026. Will this change in the future? Probably. So the concepts that we're talking about today are way more important than the tools or the prompts or the integrations because the tech's going to change, the models are going to change, but the concepts of identifying a painful problem for real businesses, proving the point, then systematizing it, then applying AI and automation, that road map won't change.
And so the big reveal here is the model that we're doing is we're creating an AI-enabled service. And even tech-heavy uh, you know, startup incubators like Y Combinator have been calling for AI-enabled services and AI-enabled agencies to join their ranks because they see that services have an enormous budget in the corporate world much more than tech directly and software. They're able now to garner kind of SaaS-like profit margins without hiring a whole bunch of people. And that used to be the thing of like, gosh, service is such a drag because I have so many clients and I have so many team members and the margins are really bad. All that kind of flips around now. You can charge a lot of money. You can do it by yourself and you don't have the hassle of like directly managing clients because everything can be automated. I won't say that everything's going to be automated from day one. You are going to get on sales calls and interact with clients directly and that's great. Uh, but the the gold at the end of the rainbow is a lot of this can be automated and we're going to get you well on your way to that today.
Okay. So the agency that we're building is uh an ads agency and we're choosing ads because it ticks a couple of boxes. Uh, one is something that people already pay money for. Uh, two is it's easily quantifiable. You know, in something like Meta or AdWords or YouTube ads or TikTok ads or LinkedIn ads, doesn't really matter. And you could swap ads for paid acquisition or LinkedIn ghostwriting or whatever the B2B services. But what I would suggest is pick something that is easy to quantify. So you can show a client, hey, we helped you do this thing and this is the financial result you got and you can do it pretty quickly. I would I would like contrast this with something like SEO or ghostwriting. That would be pretty hard because you can't really quantify in the short term something like SEO services. And I know cuz we've spent an enormous amount of time, money, and energy on SEO at my business, Castos. And it's tough to quantify the return on investment. But if you hire this agency that we're going to be setting up here to do paid acquisition for you, you can see, oh, our cost to acquire a customer uh via Meta ads in that 30 days since we hired them went down from $8 to $5. Wow, that's really great. Like the return on investment of hiring this AI-enabled agency that we're building is really clear and you can do it in a short window.
Okay, so you have this entire free repo here. You just come in here, point your agent at this repo, and you'll be like well on your way to all of the things you need to do to service clients uh with the Hermes agent. And we'll set all this up in just a minute and walk you through exactly how to set it up, how to get live, and how to start taking on your first client. Okay?
And so the stack is this. So Hermes is going to be the agent. The business model is we're selling ad services. So, we're an ad agency specifically on Meta just because I think there's the most people on Meta. There's also kind of the most opportunity I think on Meta. So, that's Facebook, WhatsApp, Instagram as of now. And so, we're going to be using a tool called Foreplay. Uh, and so Foreplay is just a giant uh kind of repository and we've created a swipe file of existing ads for our brand and for other brands. I'm going to walk through all this in just a minute. Uh, we're going to have clients fill out a bit of information that you can imagine this would be totally automatable. Just send them a link with a form and that'll pipe into our agent and we're going to collect information like this. So who's the ICP? Who's the ideal customer profile? Who are the competitors? Right? So in the example we're going to show the client is going to be Liquid Death. Uh, and I have no affiliation with Liquid Death other than I love the brand. So this is not sponsored by them. I'm not affiliated with them. I'm not marketing for them. I just want to use them as a brand because I think a lot of folks can relate to Liquid Death. It's a carbonated beverage for those of you who haven't seen them. So, their competition would be kind of other water brands, right? So, talk about who their ICP is. I'm going to show you some of this in just a minute. Who their competitors are. Uh, what kind of ad concepts do we want to come up with? And I'm going to show you what Hermes has come up with for us in this Liquid Death uh scenario. Uh, we're going to write some landing page copy because in the world of ads, there's like who am I showing ads to? What's the ad like the creative and the copy and the targeting and stuff like that? And then where am I sending them uh to like a page to convert hopefully? And then tell tell me what's going on with these ads. What does the data say both in the network and on the site? And so we're have to put those together and deliver these to the client in a really uh streamlined way.
Okay. And then you are going to be over here doing some of the work. And this is going to change over time. This is a spectrum, right? At the beginning, you're going to be doing a lot of work cuz you got to even with the starter repo I'm giving you, you're going to build a fair bit of this, right? Um, you're going to do sales for sure. You're going to kind of do the taste uh validation of this, right? It's like, hey, does this look right? Is this accurate? What do I need to tweak about the whole process and prompting and systems that we're giving you as a starting point today? Uh, you're going to be making that kind of subjective final call at the end. You're going to be managing client relationships because sure enough, even with AI doing the work of all the research and all the analysis and all the reporting and a lot of the communication, you at the end of the day want to own that relationship with the client. That's really important. So, don't ever think you're going to give that up. Uh, and then accountability, right? The hardest part of business is saying no to everything else that you can do. Especially now, like AI lets us do so much, right? And the hardest part is just saying, I'm I'm not starting a TikTok channel. I'm not starting some drop shipping thing. I'm not starting this weird, you know, directory thing over here. I'm just running a solid B2B service business. And I'm going to make my million bucks and the margins on it are going to be like 80%. Even with a big old huge spend to Anthropic or Open Router, whoever we choose to power this thing, right? So, um, that's it, right? So, Hermes is going to draft the client intake. We're going to review and send and send it to the client. Cool.
Okay. And let's get started. Now that we've identified the business problem that we're going to solve, the kind of structure of the business, let's go ahead and build the tech uh clone the repo and get started with our agent. And to do this, like you have to when you're running something like Hermes or OpenClaw, you have to have a place for your agent to run and you can run it uh on your regular computer. You can get a Mac Mini. But I got to tell you, like I was traveling for work. It's probably 2 months ago now. And power, we lost power at our house. And so I was without my agent for like 4 days traveling for work when I really kind of needed it. And ever since then, I've thought like, gosh, I want my agent to run in the cloud because then it's always on. It's managed by someone else. If something happens, I can get on the phone or chat service with them and be like, yo, you know, my my agent's down. I need your help backing things up. Um, and that's where the sponsor for today's video comes in, which is Hostinger. And Hostinger offers a really convenient one-click way for you to go from just an idea like this to having a live agent running. And we're going to walk walk through the entire process step by step. So, even if you've never installed or used an AI agent like OpenClaw or Hermes, Hostinger can take care of you. They do both really easily. Starts at about $9 a month, so it's pretty easy. I already have an account and I'm logged in here. So, from this page, and I'll have uh this link in the description below, you can just click this and just get started straight away. And there's a coupon code that'll give you a discount as well. Cool.
So, we'll just say deploy. Just going to open up our account. If you don't have an account yet, it'll ask you to create an account. And you can say for how long you want to um to create this VPS, virtual private server. It's a machine in a cloud in the cloud that is running at a data center that Hostinger kind of owns and manages. And you can do it for one month or 12 months or 24 months. For this one, because I'm going to use this for a real business just like this, I'm going to pick 12 months. And uh I get a discount here. So it's $9.99 a month, which is pretty awesome. And uh you can pick things like auto backup if you want. Pick the location that's physically closest to you. So I'm going to pick this one, Boston 2, cuz I'm in New England, so this is closest to me. And then you can enter a coupon code here. And so coupon code is in the description, but it's just my name, Craig Hewitt, all one word. Uh drop it in here, hit apply, and then continue. It'll take you to the checkout area where it already has my credit card stored in here because I'm already a Hostinger customer. Uh, but just enter your credit card information here and hit submit payment and you're all set. It's going to provision your server and going to take you to the place to activate and install your Hermes agent. And so you'll land on a page that looks something like this. And I go over here to the left and I say VPS because that's kind of where uh our agent is going to live. And I'm going to hit setup. Uh, it's chosen our server location. Same as before. I'll confirm that and just say yep. For operating system, we're going to pick Ubuntu. And we'll pick the latest one here. That's great. Uh, secure your VPS. Create a root password and SSH key. I'm going to create a root password. If you're a little more technical and you want to create a SSH key, that's cool. But I'm going to show you in a minute a really easy way to access the terminal inside Hostinger. you don't have to mess with SSH keys if you don't want to, but it's the easiest, most secure way if you're a little more technical. Uh, we're going to skip uh we'll enable the malware scanner because it's free and we want to use Docker because uh I think it's the easiest way to uh kind of cordon off areas of the machine that you're getting just for certain uh practices. And so you could have a machine that has several different Hermes agents running on it, which is pretty cool.
Okay, so now we're just setting up the machine. It's installing a bunch of stuff like the operating system. And after that, we'll go into Docker and we'll install the Hermes agent and we'll start using it. While that's installing, I want to show you around Foreplay a little bit. Uh, this is a uh library of ads. And they're getting most of these from either Google AdWords or TikTok or Meta, so Facebook and Instagram. And so Foreplay has a free 7-day trial, which I'm on right now, but we'll be using the API here a fair bit. Um, and uh, in the meantime, we're just going to be coming in here to look at our swipe file, right? So, you have swipe files for all of these here. And we have ads. So, what I would do is I would come in here and search in the discovery area for the keyword that we want to search for. So the keyword like is Liquid Death as an example, right? So just substitute whatever brand you want to search for either as like your customer. And so Liquid Death is our customer, fictitious customer. Either I'm not associated with them, not working with them. I just like the brand and and want to use it or the competitors. And so Liquid Death would be the customer, but a competitor might be Liquid IV, which is like an additive that you add to like drinks to give you electrolytes and stuff. And so you just come in here and like add these that you like to the swipe file. So I'm going to add a couple of these to the swipe file. That's cool. Uh, and then if I come over here to the swipe file, I can see that I have several different things going on already. So from Liquid IV, I have some of these. From Lollipop, which is another sparkling water brand, from Waterlue, and from Liquid Death itself. Cool.
Okay, so that's the ads and the ad intelligence. If we go back to kind of how all of this is playing together, Foreplay is the data intelligence source that's coming into both the ICP and the competitor analysis. Pretty cool. Uh, it's also going to provide the kind of guts for the ad copy. And so if we go over to our GitHub repo, which again once we get our Hermes agent set up, we're just going to point it at this and say like, hey, I want to do this. Um, we can look at clients and we have Liquid Death as an example here. And so again, you're going to be able to say like, hey, this is the example of what the input would look like and the outputs that we create from our agent. And if you go into here to inputs, you can see like we created brand voice. Uh, so like the voice, like if you know Liquid Death, it's like a really kind of biker dude kind of brand, right? Uh, so phrases to use, they're really um kind of passionate, right? And polarizing about the language they use. They're never going to say something like self-care or wellness. So So like if you've used a lot of AI, you know, you got to like put these guardrails in and say like, "Hey, don't do this and do do this." So, uh, that's brand voice. You'd want to make this a little more robust if you wanted, but this is a good starting point. Uh, client intake, right? So, like who who is this customer, right? You could get this from public sources, but you also would want the customer's input or clarification on this as well. Um, competitor list, right? Who are the competitors to the brand that I'm representing? Uh, for our water, for our Liquid Death example, we have these. That's cool. And um for offer like what is the value that this thing provides to the people that they sell it to. Cool.
Okay. So that's the input, right? This is the context that we're giving to our agent. And then there's like outputs like what is our Hermes agent going to do from understanding all this and then outputting and connecting to our ad platform. So it's going to have like uh we'll take a look at competitor analysis, right? How does my customer compare to the rest of the industry? Cool. So, uh, the Hermes agent already created all this. Uh, ICP, who is the ideal customer? Uh, sells water as an identity. Job to be done is not hydration. It is I want a drink in my hand that doesn't make me look like a wellness influencer or a guy who still drinks Bud Light. Like, yeah, that's really accurate. Okay. So this is um this is telling us a lot about in the going into the creative process like how are we going to narrow in from just the world of water to like a person. So that's what is going to be created by our Hermes agent. We're just going to point Hermes at this repo and just say like hey go go do this right and create the ways to go do this. Okay so we're ready here. Our machine is all spun up. We can SSH into this if we want. if we gave it a key. SSH is kind of a connection between your computer and the server in the cloud that this is running on. We're not going to do that because there's a way easier way to do this.
Okay, so we have Ubuntu running. We have Docker, which is awesome. Docker again is a way to just kind of cordon off stuff that's happening on that server. And now we're going to install the Hermes agent. So you can compose this manually from a link. So, if you wanted, you could do like Hermes agent.new research, right? Um Oh, they have a new page. Cool. So, you could just install it from this link. That's fine. You could do that. That's totally fine. But with Hostinger, it's like super easy. You just go one-click deploy. Uh, you come up here and just type Hermes. And I just want to install this, right? It's going to give you a username and password. And you're just going to say deploy. That's literally it. So you can see inside Docker, which again, which is a container, literally like a container, an area of this machine, this VPS that you created, I'm installing the Hermes agent and it's configuring it. It's doing all the security stuff for you kind of automatically because that's the thing with like all these open-source agent systems, OpenClaw or whatever, if you don't know what you're doing, you could set it up really wrong and for me again, like I have a Mac Mini that's sitting right over there and I never use it because I just worry that like we're going to lose power or the dog's going to come by and pull the plug or whatever and I'm going to be without my agent. So like I like running my agents in the cloud. So I can call somebody and they can get on the horn and fix this if we need to.
Okay, so this is going to take like 2 minutes to install and set up. I want to show you two things here real quick. So click the quick start and it's going to come down here and say accessing Hermes agent. So you're going to want to do two things once this gets installed and we open up the terminal. One is this first command to change directory. So change the location in the computer that we're working. So we're just going to copy this and we're going to come over here and this is all set. That's great. So we can open the terminal and here we're in the root of this server. We're going to paste in this command that we just had, right? Uh, and say, and by the way, if you've ever worked with like terminal, you know how great it is to just click a button and you're instantly in the terminal. Uh, and you don't have to worry about SSH keys or passwords or anything. Okay, so I messed up. I typed the wrong thing because this little placeholder here at the end is the identifier of your uh agent. And so you want to have this plus the last little bit from here. And so in my case, it's NYJB. So, I messed up NYJB. Cool. So, now I'm in the Remember, we talked about Docker being kind of like an area. Uh, so now I'm in that area. And now I'm going to copy and paste exactly this. And this is just a Docker command to activate the agent. Every time you open up the terminal like this, you'll do these two things. And so now uh we're in the place and in the folder where um Hermes lives and Docker Docker is activated. We can just type Hermes and it's going to open up the interactive terminal uh to run setup. Do I want to run setup? Yes. So we'll do quick setup inference provider. What are we going to do? Uh, so you can use a bunch of different models here. I'm going to use Open Router just because it gives me a lot of options, but I really also like running OpenAI Codex. So, if you have a paid ChatGPT plan, uh ChatGPT 5.5 especially works really well with Hermes, I found. Uh, I I just like a little bit cheaper model. So, I'm going to use Open Router and Kimmy K2.6 as of today. Again, the models are going to change. So, I'll say Open Router uh and we'll go over to openrouter.com. And if you don't have an account, get an account. Add a couple of keys in here. I already have uh a handful of credits. Uh, and we'll just go to API key, create a new key, and we'll call this Hermes agency bot. And we'll say create. And we'll copy this. I'll disable this key. Don't worry, you can't go you can't go use this. And we'll paste this in here. Cool.
And it's saying, okay, Open Router uh, as the name implies, is just a router. You can have access to a bunch of different models here. So, if we go back to this, we'll close this out, and we'll go to models. Uh, you can you can use so many different models, right? And so, like this is newest, but like maybe most popular, Kim 2.6, really good model, Claude. You can use uh Open Router for all of your AI needs if you want. Probably like the the subscriptions like a Codex or AI OpenAI or a Anthropic Claude subscription are cheaper on a per API credit. But if you're going to change models and model providers a lot, Open Router is a really easy way to do that because you just have one API key and you pay them and then they get a deal on tokens and that's how they make money. Um, but anyway, you have access to all the different models through Open Router which is pretty cool. So we're using this model here. So, I'm going to choose this and we're gonna say, do you want to set up messaging? Now, um, I'm going to skip this just because this is like the next step, but what I would suggest is for you to set up uh either Telegram or Discord to manage your agent because it's just the easiest way for you to communicate with your agent. Uh, unless you want to manage it like in a custom dashboard thing here. Uh, but I'm going to say skip this for now cuz we're just going to work in the terminal interface, but you probably will want an interface like Telegram. I think it's the easiest typically. And to do that, you just go in Telegram and create a bot and give it a special key and then those two are paired up. But we're going to skip this for now. Um, and you can always come back and do a Hermes setup gateway to set up messaging. Cool.
Okay. So, we can just uh do I want to launch Hermes chat now? I'm going to say yep, I sure do. And so, Hermes is starting here in the terminal. So, like it's on this machine. It's running. And in just a second, we'll be able to chat with it. Okay, cool. So, it's already running. And so, now what we can do is just come over here and grab this repo link that I gave you. Uh, it's going to be in the description below. And come back to our Hermes agent and say, and say something like this, right? I want to implement this one-person agency, clone this repo and set this up so we can start running an AI agency and hit go. Okay. And that's all that it takes to what we what have we done so far, right? Like we designed the business model. We identified the expensive problem that we're going to solve. We uh are have like all of our tech set up and Foreplay again like you have Foreplay here. You'll need an API key. Just grab this once you have a paid account uh to get all the intelligence of the ads that you're running and that uh your competitors are running and you'll be able to just drop these in here, get the API key and have your Hermes agent talk to this. And this is a good place to kind of segue. Uh, the reason I want an always-on agent like a Hermes as opposed to using CodeX or uh Codex even is I want this thing to be always able to look at what's going on in my ad manager in the platforms with reporting and things like that. And so uh having something that's always on that's just as if not more configurable than a CodeX or CodeX or Codex is really important to me 100% it's going to be more setup work for you upfront to do all the stuff to clone this repo and install it and do all the stuff here. Um, but in the end, you're going to be left with a custom solution to solve the needs of your business for your customers. And so, I think while it's a little more work upfront to do this, um, to have a like always AI agent, you need to pay something like Hostinger to host this agent for you, it being always on just gives you so much more ability to deliver value to your customers. That's how I look at it. Uh, so it's asking us, do we want to uh do this? I'm going to say yep. Uh, allow for the session. And so what it's doing now is just copying all the data from our repo here. And you're off.
Okay. And so a little big picture again. We uh have our agent. It's running in the cloud. We have Foreplay as our uh data source for ads. We have the client intake. And so you would, your agent will suggest, hey, I need a way to uh to intake data from our customers and it'll build a form for you. It will process what's going on there and create all of these deliverables for you. And again, all of that is in this repo. You just need to tell your Hermes agent to do that. And that's exactly what's happened here.
Okay. And so I hope this is helpful. I hope you're both inspired and motivated and focused to go build that million-dollar business. Like I 100% know you can. And I can't wait to see how it goes for you. Please drop a comment in below with uh your biggest takeaway from this video and the thing you're going to do today to implement this to make uh as hyperbolic as it sounds like your dreams come true cuz like take it from somebody who's doing it. Having a million-dollar business helps you sleep really good at night and gives you a lot of options in your life. And I hope that this helps you get on that path and takes you one step closer.