Claude helped me earn $90k in just 4 months

Speaker A: Welcome to the AI Paycheck Podcast. Sponsorships and advertisers are welcome on the AI Paycheck Podcast. Connect your brand with our audience. The AI Paycheck Podcast is for informational purposes only and does not provide financial, investment or legal advice. Please consult, uh, a professional before making decisions based on our content. Okay, let's unpack this because today you and I are going on a deep dive into a, well, a very specific, very fascinating blueprint.

Speaker B: Oh, it really is. It's a completely different way of looking at things.

Speaker A: Right. We are looking at an entrepreneur who used AI, um, specifically Antropics, Claude, as a core part of a business system to generate $90,000 in just four months. And you know, I know exactly what you might be thinking listening to this.

Speaker B: Yep, the alarm bells go off immediately.

Speaker A: Exactly. You're scrolling online, you see a number like 90 grand in four months, and your brain just immediately categorizes it as either a miracle or, or an outright scam.

Speaker B: Yeah, it's that classic, uh, AI prints money, clickbait that we are all just so incredibly exhausted by right now. So exhausted. But the mission of our deep dive today is to totally strip away that hype. We are going to look under the hood of this exact system to show you that this isn't magic. I mean, it is a masterclass in business systems critical thinking and hyper realistic problem solving.

Speaker A: It absolutely is. And it's vital, like absolutely vital, that we approach this not as, uh, some lottery ticket, but as an architectural blueprint.

Speaker B: Right, a blueprint.

Speaker A: Because when you actually look at the mechanics of what this entrepreneur built, the AI is not some, you know, wizard behind the curtain doing all the work while the human sleeps on a pile of cash.

Speaker B: God, wouldn't that be nice though?

Speaker A: I mean, sure, but it's a fantasy. The AI is a highly specialized tool and it's being wielded by someone who fundamentally understands how businesses actually operate, where their bottlenecks are, and crucially, what people will actually pay for. Yeah, we're moving past that whole novelty phase of AI where it was just a fun toy to play with. We are entering the utility phase and that requires a completely different mindset. So we're going to start right at that headline number, $90,000 in four months. Because it's the hook, right? It's what gets the attention.

Speaker B: Of course it is.

Speaker A: But let's immediately break down what that number actually represents out in the real world. We need to establish some trust with you right out of the gate here. That $90,000, that was revenue over a four month period.

Speaker B: Yes. Revenue, not profit.

Speaker A: Right. It was not pure take home profit that you just transferred to your personal savings account and go buy a sports car with.

Speaker B: What's fascinating here is human psychology. Honestly, because we have this deep evolutionary desire for the silver bullet, we desperately want the result without any of the friction.

Speaker A: Oh, 100%.

Speaker B: So when a new technology emerges, especially one as seemingly magical as generative AI, we project all our financial desires right onto it. We want to believe the tool itself can just generate wealth autonomously.

Speaker A: But that's a massive attribution error, isn't it? Because Claude didn't make $90,000.

Speaker B: No, Claude did not identify a gap in the market. Claude didn't, you know, hop on a zoom call and negotiate with a client.

Speaker A: Could you imagine just an AI avatar arguing over a monthly retainer?

Speaker B: Right. I mean, Claude didn't close a sale, and it certainly didn't manage the sheer anxiety of running a new venture. The human business model produced the revenue. Claude was simply a lever that the human pulled.

Speaker A: Just a lever to execute specific parts more efficiently.

Speaker B: Exactly. So if you want to replicate this, you shouldn't try to copy the $90,000 outcome. You have to copy the operating principles that made the system work in the first place.

Speaker A: I look at these claims and to me it's exactly like a weight loss commercial. You know those ones where the person lost £50 in a month?

Speaker B: Oh, yeah. And then at the very bottom of the screen in that microscopic transparent text.

Speaker A: Yes. It says, results are not typical, diet and exercise required. It drives me crazy how quick we are to attribute all the success to the software. People say Claude made me 90 grand instead of saying I worked my tail off. I built a real business and I used Claude to help me do it faster.

Speaker B: And we also have to talk about the hidden costs behind that top line revenue.

Speaker A: Oh, uh, the software sprawl.

Speaker B: Yes, the sprawl is real. You've got the AI subscription, sure, like Claude Pro. But then you need a CRM to track leads, website hosting, invoicing tools that take a percentage of every transaction. Maybe Zapier to connect it.

Speaker A: All right, all those twenty and fifty dollars monthly fees just eat away at the margins. Yeah, plus the biggest hidden cost is the founder's own time. Earning 90 grand while staring at a screen for 80 hours a week is not the dream they sell on Twitter.

Speaker B: Exactly. So if the software didn't magically create the business, we have to look at what actually did. And it starts at ground zero for any business.

Speaker A: The problem? Yes, let's get into the problem. Because the initial issue this entrepreneur was trying to solve wasn't actually an AI problem at all, was it? It was a speed problem.

Speaker B: Speed is the ultimate bottleneck. Think about a small business, right? Say a boutique marketing agency or an independent consultant. The owner has a million ideas. They want to launch a targeted campaign, draft complex proposals, document their workflows so they can finally hire an assistant trying

Speaker A: to podcast into social clips all that stuff.

Speaker B: Exactly. But they can't execute on any of it because they don't have a massive corporate team. They are completely constrained by the physical hours in the day. They have the vision, but they lack the velocity.

Speaker A: So this entrepreneur saw that gap and realized he could use AI to radically reduce the manual labor behind all those execution heavy activities.

Speaker B: Yep, he offered high speed, high quality

Speaker A: execution, which highlights what is probably the single biggest mistake people make in the AI space right now. Everyone asks, what can I sell using AI? Instead of asking the only question that matters, which is who has a painful problem I can solve?

Speaker B: Yes, people want the tool to be the business.

Speaker A: It's like buying a ridiculously expensive laser guided diamond tipped saw. Before you even know if you're trying to build a birdhouse or a skyscraper, you're just wandering around a construction site holding this fancy saw, hoping someone will pay you to cut something.

Speaker B: That analogy is so spot on. AI efficiency does not automatically equal business value. I mean, you can be the most efficient person in the world at, uh, generating thousands of AI blog posts, but if nobody wants to read them, your efficiency is economically worthless.

Speaker A: You're just highly efficient at making garbage, right?

Speaker B: True value only comes from solving something that matters to a paying client.

Speaker A: And the framework gives us three very specific criteria for a good problem to solve. These are the filters you have to run your ideas through. First, it has to happen frequently. Second, it has to cost the customer time or money. And third, the customer must already want it solved.

Speaker B: Let's break those down, because they are the difference between a side hobby and a real business. If it doesn't happen frequently, say it's an annual report they only file once a year. They aren't going to hire you on a recurring retainer.

Speaker A: They'll just suffer through it for an afternoon and move on.

Speaker B: Exactly. And if it doesn't cost them significant time or money, there's no ROI for them. Why pay you 500 bucks to solve a problem that only costs them $50 a month in lost productivity?

Speaker A: The math just doesn't work. But that third one that the customer must already want it Solved. That feels like a trap a lot of smart people fall into. Let me guess, it's because trying m to convince someone they have a problem is just basically impossible.

Speaker B: Oh, it's incredibly expensive and frustrating. You do not want to be in the business of educating a market or changing minds. You want to satisfy existing demand, step directly into the path of a budget that already exists.

Speaker A: You want the person who is currently sitting there saying, I am so sick of doing this task. I would gladly pay someone a thousand bucks to just make it go away.

Speaker B: Exactly. And once you train your brain to look for these, they're everywhere. B2B Sales Research is huge. Sales reps spend hours digging through LinkedIn before a pitch or a lead qualification. You know, sorting through an overflowing inbox to find the serious buyers or workflow documentation.

Speaker A: Every small business desperately needs standard operating procedures. But nobody wants to write a 30 page manual, right?

Speaker B: Highly frequent, costly problems.

Speaker A: Okay, so once you find that painful problem, how do you actually get an AI like Claude to help you solve it reliably? Because this requires a massive shift, we have to completely abandon amateur prompting and transition into, like, professional delegation.

Speaker B: Let's get in the weeds on this. The amateur prompt is what 99% of people do.

Speaker A: Oh yeah, they log in, they type give me 10 business ideas, hit enter and get a robotic wall of text.

Speaker B: And then they immediately conclude that AI is a fad. It's the equivalent of walking up to a random stranger on the street, handing them a pen and saying, write a

Speaker A: marketing plan with no context about your company, your voice, your budget, nothing.

Speaker B: Of course. It's terrible. These large language models are fundamentally prediction engines. They calculate the statistical probability of the next word. If your prompt is vague, it gravitates toward the most average, generic middle of the road phrasing possible. It regresses to the mean.

Speaker A: So by providing context, you are essentially narrowing down the universe of possibilities. So the AI can predict highly specific, relevant words.

Speaker B: Precisely. The entrepreneur in this 90 grand blueprint, meticulously defines the context of the objective, the specific audience, the constraints and the desired format. Plus, he provides example.

Speaker A: You're building a constrained track for the AI to run on, and there's a specific eight part prom framework he uses just to find the business operations to automate. It's brilliant.

Speaker B: Yes, the role assignment, Right?

Speaker A: It starts with act as an expert market research assistant. Then step by step, here is my business context. Here's the target customer, here's the specific problem, here's a current manual process.

Speaker B: And he asks for structured output.

Speaker A: Yeah, he tells it exactly what feels to output the current process, the AI process, human review, needed time saved, risks, and the next experiment.

Speaker B: Anthropic's own internal guidelines actually align perfectly with this. They tell advanced users to use highly structured prompts, often with XML tags, so the AI doesn't mix up the instructions with the raw data.

Speaker A: But the best part is the rule he includes at the very end. He explicitly tells Claude, do not assume that AI should be used. Recommend AI only where it genuinely improves

Speaker B: speed or quality, building skepticism right into the machine. It's crucial.

Speaker A: But you know, I'm looking at this incredibly detailed structured prompt and honestly, it doesn't sound like coding. It sounds exactly like the pedantic, overly detailed emails I used to send to my interns on their first day when I needed research done.

Speaker B: But that is exactly the right mental model. A prompt is literally a job description for a digital intern. If you give a human intern a terrible brief, you get terrible work.

Speaker A: You'll spend hours fixing it.

Speaker B: Exactly. The quality of the first draft, whether from a college student or an AI, is directly proportional to the clarity of the initial brief. You are delegating cognition.

Speaker A: So we have the problem and we know how to talk to the AI like a professional delegator. Now we need to map out the exact sequence to build this into a revenue generating machine. The blueprint outlines a five step system.

Speaker B: Let's lay out those steps.

Speaker A: Step one, find a painful problem. Step two, identify a customer already paying to solve it. Step three, create a small scoped service. Step four, use AI to radically reduce delivery time. Step five, turn the repeated work into a prominent system.

Speaker B: We need to anchor this in a real example to make it concrete. Let's talk about content repurposing. It's a massive market.

Speaker A: Okay, let's do it.

Speaker B: Imagine a business coach. Let's call her Sarah. She hosts a one hour live webinar every month packed with great insights. But afterward, the recording just sits on

Speaker A: a hard Drive or YouTuber gets like 50 views, right?

Speaker B: And she knows she should be on Twitter, LinkedIn, sending newsletters, making TikToks. But she doesn't have the time to rewatch an hour long video, transcribe it, chop it up and rewrite it for five platforms.

Speaker A: Because editing video takes hours. And staring at a blank page trying to write a LinkedIn post about something you said three weeks ago is just soul crushing. That's a deeply painful problem.

Speaker B: And she's probably already spending money on marketing, so she has a budget to solve it.

Speaker A: Perfect. So you step in with step three, your scoped offer, you tell Sarah, give me the raw video file Within 48 hours, I will return 20 distinct assets, five LinkedIn posts, three emails, five short video scripts, a blog outline and FAQ and internal sales points.

Speaker B: Now, if a human did that manually, just brute forcing it, that might take three full days of work. Transcribing, reading, pulling quotes, drafting it. Exhausting.

Speaker A: If you charged $1,000 for that, but it took 30 hours, you're making what, 33 bucks an hour? That's not scalable.

Speaker B: But step four is where the magic happens. You use AI to radically reduce delivery time. You use an automated transcription tool, then feed the text into Claude.

Speaker A: But chunked, right? Not just dumping it, all in exactly.

Speaker B: Highly structured prompts. One prompt just to extract themes, another prompt that says, act as an expert copywriter, take theme A, use the exact tone from the transcript, and draft a 300 word email.

Speaker A: Okay, but this is where you have to be super careful about hallucinations. What if the AI just makes up an inspiring quote that Sarah never actually said?

Speaker B: That is why your prompt architecture has to be so rigid. You use strict grounding instructions. You explicitly tell Claude. You may only use direct quotes found in the transcript. Do not invent or infer anything.

Speaker A: Got it. So the AI does the heavy lifting and then you step in to review.

Speaker B: Yes, the human reviews every piece, edits, polishes, and ensures accuracy. But because the AI did the drafting, it takes three hours instead of three days.

Speaker A: Here's where it gets really interesting, Sarah. The client paying a thousand dollars probably doesn't care at all if you used AI.

Speaker B: Not even a little bit.

Speaker A: It's exactly like going to a high end restaurant. You're paying for an exquisite meal. You don't call the waiter over and demand to know what brand of oven the chef used to roast the duck. You just care that the duck tastes incredible and arrives hot. The customer isn't buying Claude, they're buying the finished outcome.

Speaker B: That is such a profound realization. It shifts everything. If you pitch, hey, I use AI to make content. You commoditize yourself, you're selling the tool.

Speaker A: And they'll just say, well, I can buy ChatGPT for 20 bucks. Why do I need you?

Speaker B: Exactly. But if you sell the outcome, I guarantee a month's worth of perfectly on brand marketing assets in your inbox every Tuesday. You are invaluable. You're selling return time and peace of mind.

Speaker A: Which leads perfectly into step five. Yeah. Turning that work into a system. You do this for Sarah, then John, then Lisa, and you See the patterns? This opens the door to the Internet's favorite buzzword, passive income.

Speaker B: Oh boy.

Speaker A: Right? Because people here systemize and immediately envision sleeping in a hammock and Bali while stripe notifications ping on their phone. But the blueprint heavily cautions against this.

Speaker B: This raises an important question about the reality of entrepreneurship. We need to be crystal clear. A service business is inherently not passive.

Speaker A: Wait, really? But isn't the whole promise of AI that I can just sit on a beach while my laptop prints money?

Speaker B: It's a total myth. A service business requires active customer acquisition, pipeline management, client onboarding and intense manual quality control. If you try to make it purely passive with some zapier automations, your quality will drop to zero and you'll be fired in a month.

Speaker A: So we have to completely redefine passive. It doesn't mean push a button and never work again. It means building a digital asset that generates revenue without a one to one ratio of your manual labor to dollars earned.

Speaker B: Yes, there is a progression of scale. You start manually to understand the nuance, then you document the process. Then you automate pieces with AI. And finally, when the process is battle tested, you package what repeats.

Speaker A: Like if you've done this repurposing service for months, you haven't just made service revenue, you've created a library of prompts that actually work. Quality checklists, a ah, notion operating system.

Speaker B: And that is what you package. You sell a master prompt library or a content engine notion template. That is where you find passive leverage.

Speaker A: So the passive income is the end stage of extreme systemization. Not the starting point.

Speaker B: Exactly. It's the reward for months in the trenches doing hard service work, fixing broken prompts and learning what the market values. And even then, markets change, AI models update. You're never completely off the clock.

Speaker A: So if you're not on the beach, you're in the trenches. And to get those customers, you have to do the like work. Let's talk about weaponizing AI for research and sales without becoming an AI spammer.

Speaker B: Let's start with research again. It's about constraints. Don't ask Claude to research the marketing industry. That's useless.

Speaker A: Right?

Speaker B: Ask. We are targeting independent B2B consultants under 20 employees. Identify five recurring operational problems related to client onboarding that create create measurable time costs.

Speaker A: But the most crucial part of research isn't just validation. It's asking the AI to actively attack your idea. Like play the role of a skeptical investor. What would make this fail? Why is that so important?

Speaker B: Confirmation bias. It's the silent killer. Of all new businesses, we only seek information that validates our brilliance and ignore the red flags.

Speaker A: If the AI is just a yes man, you'll march right off a financial cliff.

Speaker B: Exactly. Forcing it to play devil's advocate. Stress tests the business before you risk a dollar.

Speaker A: Okay, so you stress test the idea now. You do customer discovery. You talk to real humans. You use AI to draft interview questions. But you have to avoid leading the witness. You focus on past behavior, not future promises.

Speaker B: How do you handle this process now? Or what does this bottleneck cost you?

Speaker A: Because it is incredibly dangerous to ask, would you buy my service? It's just like asking a friend if they like your terrible new haircut. They're going to lie to be nice.

Speaker B: The polite lie is deadly to an entrepreneur. Past behavior is the only reliable predictor of future behavior.

Speaker A: So true. So you script the right questions, bypass the polite lie, and then use Claude to prep for the sales call feed at their website industry notes and ask for likely objections. But what about outreach? Cold emailing. We are all drowning in generic AI spam that starts with I hope this email finds you well in these unprecedented times.

Speaker B: Oh, it's the worst. The rule here is cold emails must be research first, not volume first.

Speaker A: And the structure has to be tight. Four things. A specific observation proving you looked at their company, one relevant problem based on that observation, a clear reason for contacting them today, and a simple low friction next step.

Speaker B: No giant paragraphs of company history, no fake personalization.

Speaker A: The underlying philosophy is AI is best used to prepare the human for the interaction, not to replace the conversation entirely. It should increase relevance, not volume.

Speaker B: Precisely. Sending 10,000 garbage emails instead of a hundred just automates your rejection at scale.

Speaker A: Right? Okay, so you did it right. Closed the client, the money hit the bank. Now you have to deliver. This is the crucible. Let's walk through the delivery workflow because this is where the biggest risks are.

Speaker B: The workflow has distinct stages. Input, gathering raw materials, process transforming the data with AI quality check, output generation, feedback from the client, revision and final delivery.

Speaker A: The crucial concept here is the 80:20 rule. Use AI to automate the repetitive 80%. But you must manually review the critical 20%.

Speaker B: And that ratio isn't fixed. If an output affects money, public reputation, security, compliance, or contracts, human review must increase to 100%.

Speaker A: Which brings us to the terrifying reality of AI hallucinations. Fluent language does not equal factual accuracy.

Speaker B: Not at all.

Speaker A: The AI will invent a mathematically impossible pricing tier with absolute confidence. Humans have to review legal language line by line. It's like having a highly confident, incredibly fast intern who sometimes just makes up facts with a straight face. You'd never let them sign a contract without checking it.

Speaker B: The confident intern is a perfect metaphor. And beyond accuracy, there's a massive data security concern here. You cannot blindly paste sensitive client info into a public LLM.

Speaker A: Wait, really unpack that? Because I think a lot of people just paste whatever into the free chatgpt window.

Speaker B: They do. And it's so dangerous. Many free versions log user inputs to train future models. If you paste a client's secret product launch plan, it's in the system. A competitor might prompt something similar months later and see fragments of your client's data.

Speaker A: Oh wow, you could be breaching an NDA without even realizing it.

Speaker B: Exactly. You have to understand the usage policies. Anthropic, for instance, has enterprise tiers or API access where they explicitly state they do not train on your data. Professionals must configure these controls correctly.

Speaker A: So AI applied to a bad or insecure process just makes it fail faster and more spectacularly. Quality control is your moat.

Speaker B: Your quality control protocols are the brakes on the car.

Speaker A: Okay, to manage all this safely, you need a tech stack. The blueprint breaks this into a six layer business stack Layer one Research Layer two Strategy Layer three Creation Layer four Automation Layer five Delivery Layer six Measurement

Speaker B: and the AI is just a component of that stack, not the business itself.

Speaker A: Right? And you shouldn't have blind loyalty to one model. Use Claude for reasoning, Midjourney for images, GitHub Copilot for code, Zapier to connect them. But does needing all these tools just create more overwhelm for people?

Speaker B: If we connect this to the bigger picture, it explains a fundamental shift in the economy. The real skill of the future isn't just typing a clever sentence, its workflow, design and tool selection. Think of a master carpenter. He doesn't feel overwhelmed by having a hammer, a saw and a drill. He knows which tool to grab for which stage of the project.

Speaker A: Let's unpack the everyday AI tips from this blueprint to build that intuition giving context we talked about. But audience definition is huge, right?

Speaker B: Oh, totally. Writing an email to a Gen Z consumer is radically different than writing one for a 50 year old compliance office. If the AI doesn't know the audience,

Speaker A: it defaults to corporate speak and defining the format explicitly. Don't just ask for a list if you need a CSV file or a markdown table.

Speaker B: Right? Dictate the structure up front to eliminate

Speaker A: manual reformatting later, also providing examples for consistency. Don't just tell it to be funny. Paste in three of the client's past posts and say match this tone.

Speaker B: It gives the AI a specific target and identifying assumptions.

Speaker A: You can ask the AI what assumptions am I making here that might be false?

Speaker B: Yes. Or ask for outright counterarguments. Have it violently argue against your proposal from the client's perspective to reveal hidden objections.

Speaker A: That's amazing. And for facts. Instruct the AI to explicitly separate verified facts from creative suggestions in the document.

Speaker B: It helps so much with quality control.

Speaker A: And you have to save your useful prompts. Don't reinvent the wheel. Build a prompt library because when you

Speaker B: sequence those saved prompts, you create real workflows. The AI summary of a call becomes the input for the follow up email which becomes the input for the CRM notes.

Speaker A: But the golden rule. You must review the outputs yourself. Your name is on the deliverable.

Speaker B: You face the consequences, not the AI.

Speaker A: The blueprint is also brutally honest about failures. This entrepreneur tried things that did not work. Zero demand for some automated workflows, unusable outputs from certain prompts, and if a customer demanded too much bespoke customization. The AI efficiencies collapsed and destroyed the profit margins.

Speaker B: That's the gritty reality. Failures are where calibration happens. You have to break things to find where AI actually creates economic value.

Speaker A: All right, we are rounding the corner here. Let's consolidate this into an immediate action plan. A concrete business plan for the listener. It starts with a 30 minute exercise. Three columns on a piece of paper.

Speaker B: Column 1 skills I have not AI, just what you know. Writing, accounting design. Column 2 problems I intimately understand. Column 3 tasks people actively pay for.

Speaker A: Find the overlaps, identify five micro steps where AI could reduce the workload and then do not build anything yet. Don't buy a domain name. Go talk to customers first because you

Speaker B: need to understand the revenue math. Focus on unit economics. Five customers paying $1,000 is a $5,000 business. But delivery time matters. If it takes 20 hours per client, it's just a low paying second job. If AI compresses it to two hours, you have a highly profitable system.

Speaker A: So the 30 day plan. Days one through seven choose your customer group. Do past behavior interviews. Find the recurring problem. Just listen.

Speaker B: Days 8 through 14 build a small tightly scoped offer. Use CLAUDE to draft the pitch. Deliver the service completely manually to a beta client to collect brutal feedback.

Speaker A: Days 15 through 21 improve the process based on feedback. Document the steps, create reusable prompts and reduce delivery time to that 80:20 ratio

Speaker B: and days 22 through 30. Sell the improved offer, track leads, revenue, delivery time and profit. Then make a rational decision on whether to scale it or pivot.

Speaker A: And you need a one page seven line business plan Line one Customer Line two Problem Line three Current solution and why it sucks Line four AI Opportunity for speed quality Line five Exact offer Line six Price Line seven Proof you can deliver. That's it.

Speaker B: It strips away all the corporate bloat and forces you to confront the mechanics of commerce.

Speaker A: So what does this all mean? It means we are in an epidemic of people stuck in tutorial purgatory, collecting subscriptions like gym memberships, watching YouTube videos, bookmarking prompts they'll never open. They feel productive but haven't solved a single human problem.

Speaker B: It's the illusion of progress, speed of execution, and delivering a tangible result matters infinitely more than showing off an AI parlor trick.

Speaker A: A beautifully coded AI automation that doesn't relieve human pain or save a business money is a hobby. It's not a paycheck.

Speaker B: If AI can commoditize the execution of a task, then the only truly valuable skill left is your ability to accurately diagnose the human problem in the first place. Are you spending too much time learning to speak to the machine and not enough time learning to listen to the market?

Speaker A: If today's idea gave you one new way to make money with AI, follow AI Paycheck podcast. We publish one practical AI Income idea every day. Know someone trying to make extra money with AI? Send them this episode.