How to Build an AI Business (Without Guessing What to Sell)

0:00 You've been meaning to start something for about four years and every time you get close, you hit the same wall. You don't know what you would sell. You have no audience, no list, nobody to sell it

0:10 to. And you'd have to learn six things you have absolutely no interest in learning probably. Those are all real reasons in 2021 or 22. But today, if I had to start a business from absolute

0:23 zero, this is how I would use AI to do it. There are five stages between where you are right now and your first customer. And I'm going to take you through all five. And I have broken down

0:34 the AI workflows that I would use at every stage, including a few that I think most people starting businesses completely overlook. So let's start at zero. You don't even know what business

0:44 to start. Basically, this is where I think most people already use AI wrong because they open Chad GPT or Claude and go ahead and type, "Give me 20 profitable business ideas." And guess

0:57 what? AI happily gives them the same list that it gave to 10,000 other people. AI agency, drop shipping, newsletter, SAS, online course. That's not opportunity discovery. That's like

1:08 Google search in a tool that's a thousand times smarter. If I were actually starting from zero, I would do something very different. First, I would turn myself into the data set. Yep. I

1:19 would make AI interview me about my career, things that I've learned, industries that I understand, people I know, problems I've solved, things people already ask me for help with,

1:29 skills that I probably don't even think of as skills anymore. And then I would add my constraints. How much money I can invest or if none, how quickly do I need revenue? Do I want clients or thousands

1:41 of customers? Do I want to be the face of the company? How many hours can I give this? And so on and so forth. Because honestly, a great business opportunity for someone with half a

1:51 million and a software team might be a terrible idea for you and vice versa. And then I would ask AI to find intersections between three things. What I'm good at, what people need, and what

2:03 they're willing to pay for. And that gives you something much more useful than a list of trendy businesses. It basically gives you your founder market fit. And later on, I'm going to show you

2:15 the one where AI sits in your calls, in your sales calls, and rewrites your offer for you. That one completely changed how I think about all of this. I'll get to it, but before that, I want

2:27 you to take this with you. So, I'm going to make sure that you have access to all the prompts that I'm going to be using. Okay. The prompt that I would use for this one is long and it does a lot of

2:40 things that maybe some people don't think about. I'm not going to read it out loud because otherwise this is going to become a 2-hour video, but I'm going to make sure that you can screenshot it

2:49 from the screen here or here. And I will also make sure that you can grab it as a downloadable if you want to just copy paste. We will make sure all of the prompts and all of the details are in

3:01 one single document inside our founders community and you're able to go there, download it. It is a paid community, but you get 7 days for free. So, you don't have to stick around if you don't want

3:12 to. I do encourage you to check it out because there's a lot of value in there and you might decide it is worth your time. It might help you move a lot faster on this journey actually. So,

3:23 pause the video right now and copy it or get it from our community. It does come with a challenge as well. So, the link is going to be down below. Now, let's run it for a former client of ours.

3:34 Okay, I'm going to run it here on my computer so you can see how the results look like. Now, you see the recommendations are very personalized, but I still wouldn't pick anything yet

3:46 because this is still based on what I know and the market knows much more. So the second workflow that I would recommend if you want to start from zero is to mine the internet for problems.

3:58 Okay, I would turn AI into an opportunity researcher. Let's say that you're considering starting something for independent accountants or realtors or chiropractors. Okay, you can go ahead

4:09 and research Reddit threads, YouTube comments, industry forums, communities, customer reviews, competitor testimonials, and instead of reading 2,000 different conversations yourself,

4:19 AI can help you cluster them. What are people repeatedly complaining about? What takes too much time? What are they already paying someone to do? What solutions have they tried and hated? So,

4:29 basically, you get all the valuable juicy things like, I wish there was this or why doesn't somebody do that? That's where things really get interesting. Now again, I'm going to do it real time for

4:41 you, but the prompt is going to be here or here. So you can pause and copy paste it. And you see, it finds thousands of conversations, lots of complaints, many recurring problems, and some potential

4:54 opportunities. And you see, you're no longer asking AI, what business should I start? You are literally using it as a consultant and you're asking it where is there evidence of an expensive, frequent

5:05 or frustrating problem that I may be unusually qualified to solve. It's a very different question obviously. Now the third workflow is to ask AI to destroy your idea. It's something that

5:18 most people wouldn't do, right? But if you tell your favorite idea to AI and then ask it to judge it, you are going to end up with a much stronger idea. So you can ask AI to behave like an

5:31 investor, a customer, a competitor, a CFO as well as a deeply skeptical entrepreneur. And you will see what feedback you get that will be so invaluable for you to make sure you get

5:42 started on the right track. Because we all agree that AI is good at making your idea sound clever, but sometimes you needed to do the opposite. Okay. So, by the end of stage one, you don't want to

5:53 have a shiny business that will die after two months. You want a hypothesis that is worth testing. So, again, we're going to give you the prompt. You can pause, copy paste, and use it directly

6:04 yourself. Now, sure, this is already much better than you would get with a typical prompt, but honestly, it's worth almost nothing until somebody who is not you also agrees with it. And this next

6:16 stage is the one that I come back to at the end of the video. I do that every time when I work on a new project, especially after getting customers and not before them. You'll see what I mean.

6:27 Now, there is a very expensive mistake that founders make. They fall in love with the solution or the product or the service before anyone proves to care about the problem. I've done it. I've

6:39 built things that were fantastic, or at least I thought they were, and then discovered the market was significantly less excited about our genius than we were. Like for example, the time we

6:49 thought that we had invented a new phone charging stand or the time when we thought selling access to an app as a one-time fee for lifetime access was a great idea, but clients actually wanted

7:00 to pay for usage. This was 2009. So what I need you to understand is that the product is not the business. The customer wanting the product is your business. And this is why in our own

7:12 work, product market fit and idea validation always happen before we start designing the conversion system around them or the offer itself. So how do you do that? Workflow number four is to turn

7:24 the internet into your focus group. You can start with all the conversations that you've collected earlier. Now you're going to not be looking at ideas but studying the pattern in terms of the

7:35 problem. What's the deep problem coming up? Okay, you can give AI all the hundreds or thousands of relevant conversations and you can ask it what outcome are these people trying to

7:46 achieve. You can get so much deeper in understanding exactly what words they use to describe the problem, exactly what alternatives they're thinking about and this will set you up for success.

7:59 And that last one is really, really important. Okay, don't replace their language with clever marketing one. If 10 customers describe the problem as I spent my entire Sunday catching up on

8:09 admin, don't go and turn it into optimize your operational efficiency. No, your customer just wrote better copy than you did. So use that, okay? Understanding the language they

8:19 naturally use is a thousand times more powerful than inventing language in a marketing meeting. You don't have to do that. But there is still a problem. Okay? AI can analyze a market and give

8:29 you a very convincing answer. But that does not mean that the market agrees. What you really need to do is AI assisted customer interviews. Okay, this part I would not entirely delegate to

8:40 AI. Idea is to talk to potential customers. Okay, because AI can find your candidates. It can help write good non-leading questions. It can transcribe the calls, but you need to talk to them.

8:52 And after doing 10 or 15 or 20 conversations, you can also ask AI to compare every transcript and find patterns that you would probably miss manually. Trust me, we've taught this

9:02 exact approach in our community. Actually, interviewing people who have the problem and learning what they've tried, what's preventing them from solving it, and what they're hoping to

9:10 find instead is one of the best ways to make sure that you pick a business idea and you design it in a way that will not fail. You have to hear the customer, okay? Because the hesitation before they

9:23 answer, the thing they get emotional about, the throwaway sentence that they repeat five times, those are the things you need. AI can make sure the research is scalable, but AI cannot remove the

9:35 customer from customer research. Okay? You can use AI in multiple ways. Like I said, it can prepare and analyze the conversation itself, stays human, but then you can use tools like Google Meet

9:48 or Zoom or whatever you decide to hold the call on. You can use Fathom or Fireflies if you want to record the call, which I do recommend because then your AI is able to much easier get

10:01 access to your transcripts. You can use Calendarly, for example. It's completely free in the beginning to do the bookings and then you can use your LLM for the question set beforehand. Okay? And then

10:13 you can hand the transcript over for the analysis. Make sure you ask for permission before recording. By the way, now that you know how to get the customer insight truly from the

10:23 customer's mouth, we are ready to move to the next step, which is a very exciting one. We need to turn the evidence into an offer. Okay? We need to give all of this back to AI. interviews,

10:34 the reviews, the competitor analysis, the problems, the desired outcome, the objections, and ask it to help us design the smallest compelling offer that solves the most important problem. Okay?

10:44 So, you move from consulting for chiropractors or realtors to I help independent chiropractors follow up with missed inbound leads automatically. So, now you know who your service or product

10:57 is for, what problem you're solving, what outcome you're promising, and what you're actually selling. I'm not going to talk about pricing. I cannot recommend how much to charge. If you

11:06 want, I can make a dedicated video about that. It can get long. Or you can join our founders community. I have dedicated lessons about how to design your offer and how to set up your pricing. But

11:19 here's where 2026 gets particularly interesting. You don't necessarily have to spend six months building it because AI can help you create your landing page, a clickable prototype, a mockup, a

11:30 demo, a weight list, maybe even a functioning first version. Okay? It depends a lot on what your offer is and how you want to go about it. But you are able to get your MVP, so the minimum

11:41 viable product ready within hours or maybe if you're a beginner, days. And here's the prompt that you can use to come up with your MVP idea. Basically, what you want is to build enough to test

11:53 the assumption, not enough to satisfy your ego because if people don't want the small version, adding another 47 features probably is not going to save you. Now, we have something that people

12:03 might actually want, but nobody knows it exists. So, the third big step is becoming findable. This is where people usually say, "Okay, I need a website." And my answer is not quite. You need

12:17 somewhere for demand to find you and understand you. A website nobody visits is not a business asset. It's a brochure that is sitting in someone's drawer digitally speaking. So what you need is

12:29 to turn your brief into a digital storefront. And the good news is that all the work we've already done becomes context. All right? So you can give AI the customer, the problem, the offer,

12:40 the positioning, the objections, the proof, the language that customers actually use. all of that together as context and knowledge and now it can help you create your digital storefront.

12:50 It can come up with the site architecture, with the homepage, with the offer page, everything else that you need, including lead capture and the initial SEO structure. And with today's

13:00 vibe coding and AI website builders, you can basically go from brief to something functioning quite quite quickly. And you won't need to write copy or hire a developer in the beginning. We're going

13:12 to give you the prompt that you can use as well. And think about it. A few years ago, you would have lost weeks here, but today you can get a credible first version this afternoon. But there's a

13:23 part that a lot of people overlook. Okay? Before deciding what pages to create, you need to ask yourself, what would somebody type when they already have this problem, right? Because maybe

13:33 your customers are not searching for the name that you've given your category. Maybe they're searching how to respond to missed dental calls or maybe they're searching dental practice losing leads

13:44 or something completely different. I don't know. You need to figure that out. And AI can help you map it. So, how do you go about it? Basically, you need to go from problem to search intent to the

13:56 question they ask, the page content and your offer. Okay? So, your website architecture starts reflecting the customer's world rather than your internal company structure. And this now

14:06 goes beyond traditional Google search, okay? Because people increasingly discover businesses by asking LLM's questions too. So I want my business information to be extremely clear. This

14:16 is what you would tell your AI. And you would say, "This is who we are. This is what we do. This is who this is for." At the end, you need to ask it to help you write it in the words of your customer.

14:31 And in about 2 minutes, I'm going to take one 20inut conversation and turn it into a week of content across five platforms. No AI slop, I promise. The same thinking, but in different shapes.

14:42 That's coming next. But there's one more thing because everything I just showed you, you can do once in an afternoon, but doing it again next month is the part that does not happen. And next

14:54 month, you will be busy. Okay? Which is why for this video, we are partnering with Hrefs. Hrefs is the tool. Let is the worker. It is a workspace where you describe what you want in plain English

15:08 and then it goes and builds it. Not a chat that gives you advice. Something that does the job and keeps doing it. Watch this. I am telling it in one sentence to check every week whether

15:17 people can find me and what competitors rank for that I don't. I'm not building anything. I'm just describing it. And then it goes away and comes back with the finished thing. Now, you've heard of

15:28 vibe coding. I'm sure this is that but pointed at the business side instead of the software. 99 a month and it sits on an HF's plan for the data. This is not buying you the tools. This is buying you

15:41 the person who would otherwise be running them every week which right now is you at 11:00 probably during a week night. Link is going to be in the description. Thank you so much HFS for

15:53 partnering with us on this one. So what you want AI to help you with here? You wanted to help you find what people type when they have the problem. So, the site is built around their words rather than

16:03 your category name. And here you can use HF's free keyword generator. Basically, it's free. There's no sign up, no card, and it returns keyword ideas with estimated monthly volume and has a

16:14 questions report. So, what happens is you take the customer phrases from workflow 4 and put them into the keyword generator. You take everything with real volume and then you discard everything

16:26 that doesn't meet the required threshold. But being findable only captures demand that already exists. What if nobody's looking for you yet? That means you need traffic. And this is

16:37 usually the stage that scares a lot of people because they think, "Oh, I don't have an audience or I have not done this before." Fine, you don't need an audience to start. There are really two

16:48 games here. You can build attention over time or you can go and find people who already have the problem. You can choose between warm traffic and cold traffic. Each has trade-offs. Okay, an audience

16:59 builds trust but takes time while cold traffic can be switched on much faster but usually begins with less trust and can become expensive or effort heavy in the economics of the beginning. AI can

17:11 help you with both. Let's talk about this next workflow. If you're doing organic, don't ask Chad GPT or Claude, give me 30 Instagram ideas. Please don't do that. We already have something much

17:22 better. Okay, we have the customer research. So, what you do is you give AI the pains, the questions, the objections, and the desired outcome that we found in stage two. And then you ask

17:31 it, what does this customer need to understand before they could ever buy? Okay, again, full prompt is going to be available. I'm not going to read it because it's way too long and you and I

17:41 don't have time to sit here for hours just for you to wait for me to read everything. Now, from there, one customer problem can become a YouTube video, a LinkedIn post, a newsletter, a

17:52 short, a case study, a search article. And really important, not all content has the same job. Some content exists to reach people who've never heard of you. and some other content pieces are there

18:05 to deepen trust with people that are already paying attention. That distinction between growth and core content is very important especially for YouTube. Now the next workflow is to

18:17 take that one idea and turn yourself and your company into a media company. That is exactly where AI gives a tiny company something that previously required a content engine or team. Imagine I record

18:28 one 20 minute conversation. No AI nonsense, okay? Just me thinking out loud. I can extract the strongest ideas, find clips, turn one argument into a newsletter, adapt another for LinkedIn,

18:41 create shorts, draft a blog post, suggest follow-up topics based on questions that come up. So instead of creating 12 disconnected pieces of content, I create one valuable idea and

18:51 give it more surface area. That's a very different use of AI. your job becomes to think and to have the original ideas and then AI can take that and distribute. If I'm starting tomorrow and I need

19:05 customers quickly, I probably would not wait for my YouTube channel to hit 100,000 subscribers. Let's be honest, okay? I would go and find them myself. Let's say that my customer is an

19:15 independent chiropractor or a realtor or dental practice or a lawyer. Okay? AI can help build a list of relevant prospects. research each business, study their website, identify relevant

19:27 signals, and then help you create outreach that actually references something specific about that company. You know, you don't want to do the dear sir, I noticed you're very impressive

19:36 company in the industry of so and so. You need actual relevance. And then it can help you track replies, draft follow-ups, update your CRM with the correct automation. But the scale is not

19:48 the interesting part. The context is AI means personalized outreach doesn't necessarily have to become generic outreach as volume increases. What it does is research and drafting at volume

19:59 without the message becoming generic as volume increases. Now, if you want me to walk you through step by step how this workflow would look like, I can take one of the ones that we have created for

20:10 some of our clients and I can show it to you step by step. But that would need to be a separate video. So, let me know in the comments if that's of interest. If you have money to test paid traffic, AI

20:20 can unlock another bottleneck, which is creative experimentation. You can take one customer problem and generate 10 positioning angles, five hooks per angle, different visual concepts,

20:30 different landing page variants and run ads. Okay? And then once real performance data comes in, you can give that back to AI and don't just ask which ad won or which one's better. You can

20:42 say what are we learning about the customer from what won. Maybe convenience beats price. Or maybe fear of lost revenue beats desire for growth. Or maybe one customer segment converts

20:54 much much better than all the other ones. And now advertising becomes customer research as well because one customer insight can become a large number of tested messages and then the

21:04 results will teach you about the customer rather than just giving you the winner. Obviously, this one requires that you also understand at least a little bit how Meta Ads Manager works or

21:16 Google Ads Manager if that's the platform that you're going to choose. And I want you to notice what we're doing here, okay? Because every stage is now starting to feed the others, but

21:26 views don't pay you and traffic doesn't pay you and followers don't pay you. Eventually, somebody has to pay you. That's stage number five and that's convert. Okay, that's the final mistake

21:38 that I would avoid. People spend months trying to get more traffic when sometimes the problem is not traffic at all. It's what happens after someone arrives because marketing should

21:48 ultimately connect to a sales outcome rather than treating impressions or awareness as proof that the business is working. That's something that I've talked a lot about in our founder work.

21:57 So, if you're interested, our founders community is always ready to welcome you. Let's talk about how you turn your website into a salesperson, okay? Because you need to give your website an

22:09 AI layer. And no, I don't mean, hi, how can we help you today? I mean an AI assistant that actually knows your offer, your products, pricing, FAQs, case studies, qualification criteria,

22:20 common objections, and when somebody arrives at midnight and they ask, "Would this work for a three location practice?" The AI can answer based on your actual knowledge base. And not only

22:32 that, it can ask qualifying questions. It can recommend the right next step. It can capture the information, maybe even book the call. That's very different from putting a contact form on a page

22:42 and hoping that somebody fills it in. So here, obviously, you need to first of all assemble your knowledge base. And then you need to use that knowledge base together with an agent that is able to

22:53 at least communicate with people who land on your website through text. Ideally, you would be able to have also a voice agent that people who are on your website can call and get direct

23:04 feedback from on the spot. And then ideally, you can connect it to your calendar, to your booking link, if you have something like that. Now, let's say that somebody actually books. This is

23:16 where AI becomes super useful. You can have it transcribe every sales conversation and then across 20 or 50 calls, you can go ahead and get insights. You can ask what objections

23:27 keep appearing? Which problems correlate with buying? What language do customers use? Where do people lose interest? And what alternatives are they considering? Why do we win and why do we lose?

23:38 Suddenly, your sales calls are not disappearing the second that Zoom or Google Meet or whatever you use closes. They are becoming a customer intelligence database. And if you're

23:50 wondering where all of that information goes, back there into your product market fit. I told you at the start that stage two is one we would end up back into. That's it. This is why I don't

24:01 really think of these five stages as a funnel. I think of them more as a learning loop because this way sales conversations stop evaporating. Okay? And they become the best customer

24:12 research that you will ever have. This is the workflow that closes the loop. You can use any kind of recording tool you choose. You can use any LLM that ideally has connectors with that uh

24:25 recording tool and then create a way for your LLM, whether it's Claude or ChatgPT, to go and check those transcripts. And then AI can also remove a lot of admin friction around selling.

24:40 Imagine finishing a discovery call. AI already has the transcript. It can extract the requirements. It can update your CRM. It can create the meeting summary. It can draft a personalized

24:49 proposal based on what the customer actually said and prepare the follow-up email. And if you have a team, it can even create the next actions. You review, you approve, and you're done.

24:59 So, you stay in the decision mode and AI is able to do the repetitive work around the decision because between the call and you sending out the proposal, that's where deals go cold and that's where a

25:12 lot of deals end up being lost. And now finally, once we have enough data, I would give AI an entire conversion path. The number of visitors, the number of leads, the number of calls, and the

25:23 number of customers you've closed. And then ask it to investigate where are you losing people? Which source produces the best customers? Which landing page converts better? Which segment buys?

25:33 Which objections correlate with lost deals? And what should you test next? That's where AI becomes your chief of staff, your decision support system because it is able to turn a wall of

25:46 analytics into diagnosis as well as into your next test. This is where you use AI intentionally as your business partner. So look at what you just built. You started with nothing, no idea, no

26:00 audience, no website, no leads. And you use AI across five stages. To pick the opportunity, to find product market fit, to become findable, generate traffic, and convert that traffic into customers.

26:11 But I want you to notice what I didn't say. I didn't say AI chose the business for you. I didn't say AI proved the demand. I didn't say AI built the relationships. And I definitely didn't

26:20 say AI convinced someone to give you money while you did nothing. Because I think that's the wrong lesson from what's happening right now. The interesting thing is not that AI removes

26:31 the founder. It just removes a huge amount of friction around the founder because research that used to take days or weeks, you can now do in hours. Things you couldn't design, you can now

26:42 prototype. Skills that you didn't have, you can now get assistance with. Information you couldn't possibly process manually, you can now interrogate. Work that once required

26:52 several people, especially specialists, now is available to you. But judgment, taste, trust, decision- making, actually talking to customers, and making the final call, that's still you. And I

27:04 think that's a good news. All right, I hope this was helpful. I hope you can see that AI really lowered the distance between an idea and the business, between evidence and something that you

27:14 can build. And really, when the cost of trying falls, you get to test more. You can learn faster and spend less of your life building things that nobody asks for. So, if you've been waiting for the

27:24 perfect idea, the perfect skill set, the perfect website, or the perfect audience before you start, I wouldn't. Just start with stage one. Find a problem worth solving and use AI to get further

27:36 faster. Thank you so so much for watching this video. I really hope it was helpful. Like it if you did. Be sure to subscribe if you haven't done so. Share it with anyone in your circle of

27:44 friends or family or co-workers who's been thinking about starting a business and is still procrastinating even though they understand that AI can do a lot for them. Until next time, I suggest you go

27:55 ahead and watch this video here and I am looking forward to welcoming you into our communities. I'll see you there. Bye.