I'm going to be brutally honest with you here. Three months ago, I had zero income from any kind of AI side hustle. Today, I'm pulling in a consistent $3,400/month from a business most people scroll right past: an AI API reseller operation. No, I didn't train a model. No, I didn't raise venture capital. I just figured out where the margin actually lives, and then I went after it.
This is the full story. The wins, the flops, the screenshots I almost didn't take, and exactly how the numbers shake out each month. If you're thinking about starting something similar, pull up a chair. I'm about to save you about six months of trial and error.
How I Stumbled Into This Entire Thing
Here's the truth nobody talks about: I didn't have a grand vision. I was doom-scrolling Twitter at 2 AM last October, and I stumbled across a thread where some indie hacker was bragging about making $2,000/month "reselling" AI access. I almost kept scrolling. Something about the word "reselling" felt sketchy — like the kind of grifty thing you'd see in a YouTube ad.
But I'm the kind of person who can't let a weird idea go. So I started pulling the thread. What I discovered genuinely shocked me. There's this whole underground world of people who wrap existing AI infrastructure into niche-specific products and charge a markup. They're not building models. They're not even building complex software. They're solving a specific problem for a specific audience, and the audience is happy to pay for the convenience.
That was my lightbulb moment. I closed Twitter and opened a spreadsheet.
The First $0 Month (And Why It Was Important)
I'm not going to sugarcoat this. My first month, I made exactly $0. I didn't even count the trial credits I burned through testing things. I literally had zero dollars in revenue.
But here's the thing about build-in-public culture that I love: the zero months are where the real learning happens. I spent that first month doing what I'd later call "foundational faffing" — setting up infrastructure, writing landing pages nobody visited, and messaging people who had no interest in talking to me.
What I learned during that worthless month ended up being worth more than any blog post or course. I figured out which platform to anchor everything around, which niche I wanted to serve, and (most importantly) that I had absolutely no idea how to price my service. All three lessons cost me time but zero dollars, which felt like a win.
Picking the Engine: Why I Chose Global API
I tested four different underlying platforms before settling on one. I won't name the others because this isn't a comparison post and I'm not trying to dunk on anyone. What I will tell you is that the decision came down to three things: model variety, margin potential, and how the affiliate program was structured.
Global API won for one reason that I think a lot of people overlook: it gives you access to 150+ models through a single API key. That sounds like a small detail. It isn't. When you're reselling to a niche audience, you don't know in advance which model they'll want. If a customer asks me tomorrow for something specific, I don't have to scramble, sign up for a new provider, and rebuild my whole integration. I just toggle the model in my dashboard.
The affiliate structure also made it incredibly easy to start. First-order commissions sit at 15%, and recurring commissions are 8%. There's also a 10% premium tier for partners who move more volume, which I'll get into later because hitting that tier changed my monthly numbers significantly.
I want to be transparent here — I didn't pick Global API because someone paid me to. I picked it because, after running the actual calculations across four competitors, the math worked out best for the kind of slow, steady growth I was targeting.
The Niche Decision: My Ugly Spreadsheet Moment
I'm going to show you my thinking because this is the part most guides skip over.
I made a spreadsheet with five columns: niche, addressable market size, my ability to add value, competition density, and average willingness to pay. Then I scored each one out of 10. The niches I considered were:
- Content marketing agencies
- E-commerce product description writers
- Real estate agents needing listing copy
- Indie game developers
- Small law firms doing document summarization The real estate one kept scoring weirdly high. The market is enormous, willingness to pay is genuinely high (agents already spend thousands on lead generation tools), and — this is the kicker — the existing AI tools either ignore them completely or charge enterprise pricing they can't stomach. That's when I knew. Real estate was my niche. I want to pause here because I need to make something clear about the build-in-public movement: sharing my niche decision publicly was terrifying. I kept thinking, "What if someone reads this, steals my exact playbook, and beats me to market?" Then I realized something important. Execution always beats ideas. I've now told probably a hundred people exactly what I'm doing, and none of them have actually built the thing. The fear of being copied is almost always worse than the actual copying. # # Building the Actual Product (A Week of Ugly) Here's the part that made me feel like an imposter for about three days straight: I built my offering in a week, and most of it was ugly. I started with a Notion page. Notion! That's how little I thought of this initially. I told myself I'd just collect emails, learn what people actually wanted, and then build the real thing. The Notion page described three "packages" — Starter, Pro, and Agency — and it had a Typeform link for sign-ups. Within four days, I had eleven sign-ups and zero paying customers. That's when I knew I needed to actually build something. I spent the next weekend putting together a proper landing page on Framer, a Stripe integration for payments, and a simple dashboard where customers could manage their credits. None of this was revolutionary. The whole thing looked like it was built by someone who was clearly building it for the first time. But it worked, and that was all that mattered. # # Month 2: My First $847 I almost didn't share this number publicly. $847 feels embarrassingly small. Most build-in-public creators I follow only post the months where they made five figures. But I'm doing this differently on purpose. The whole point of documenting my journey is to show the actual climb. So here it is: month two, I made $847. Where did it come from? Three paying customers on the Pro plan ($199/month each) and one agency client ($250/month). The agency client found me through a real estate Facebook group where I'd been lurking for two weeks before I ever posted anything. That's the lesson — the customers don't come from your landing page alone. They come from the hours you spend in the communities your niche already lives in. I want to break down the actual economics because this is where the build-in-public transparency really kicks in. My revenue for the month was $847. My costs were:
- Global API usage fees: approximately $254
- Stripe fees: roughly $25
- Landing page and tool subscriptions: $43
- Total costs: about $322 Net profit: $525. For roughly 15 hours of work that month. That's $35/hour, which is better than my day job paid me when I started freelancing back in college. # # Month 3: The $3,400 Breakthrough (And What Changed) Okay, this is the month I want to talk about because something shifted and I want to be specific about what. Three things happened almost simultaneously:
- I crossed the volume threshold that bumped me into Global API's premium affiliate tier (the 10% rate). This alone added a few hundred dollars to my monthly numbers because I was now earning on every API call my customers made, not just the initial signups.
- I launched a "Done-For-You" tier at $499/month where I don't just give access to the API — I actually configure workflows for the agent, set up their listing templates, and handle the initial prompt engineering. Six customers signed up in the first week.
- I started posting monthly income reports publicly. I have no idea why this worked, but my sign-up rate roughly doubled. Apparently, the kind of person who wants to buy from a small AI reseller is also the kind of person who wants to see proof that the reseller isn't a scam. Let me do the full month 3 breakdown because I love looking at this: Total revenue: $3,417
- 11 Pro plan customers: $2,189
- 6 Done-For-You customers: $2,994
- Recurring affiliate commissions (8% on the renewals from customers I'd referred in earlier months): roughly $234
- (Note: the affiliate commission numbers above reflect the 8% recurring rate on prior-month customer renewals, while the Done-For-You and Pro revenue figures represent what my customers paid me directly for my own service) Wait, let me re-do that math more carefully because I want to be honest with you. The bulk of my $3,417 came from the 6 Done-For-You customers and 11 Pro customers paying me. The affiliate commissions were a smaller but meaningful chunk layered on top. Costs for the month:
- API usage: $1,142
- Stripe fees: $102
- Tools and subscriptions: $58
- Total: $1,302 Net: $2,115. That was a life-changing month for me. Not because $2,115 is some massive number, but because it proved the model scales. # # The Parts Nobody Wants to Admit Now let me share the stuff I almost left out of this post. Customer support is a real job. I had this fantasy where I'd build the system, the customers would pay, and I'd just collect money while sipping coffee. That lasted about eleven days. Real estate agents ask questions at 7 AM, 11 PM, and on Sunday afternoons. They want hand-holding. I'm not complaining because the hand-holding is part of the value I sell. But if you're thinking about doing this, block out at least 4-5 hours per week just for support, especially early on. One niche customer can ruin a week. I had a customer in month 2 who demanded a full refund, claimed my service "didn't work," and then proceeded to leave a review that I had to respond to publicly. It cost me about 8 hours of stress and at least one potential customer. This is part of the gig. You will have difficult people. You will lose sleep over them. Then you will move on. The recurring revenue is everything. I made a real mistake in my first month by chasing one-time payments instead of subscriptions. If I'd structured everything as monthly recurring from day one, my month 3 number would have been higher. Don't repeat my mistake. Get people on subscriptions from the start, even if it feels awkward. # # The Affiliate Program That Made This Possible Here's where I have to address the elephant in the room. I'm going to talk about the Global API affiliate program, and I want to be upfront that yes, I'm a member of it. But I also want to explain exactly why I joined, because I think my reasoning applies to anyone reading this. When I started, I needed three things:
- A reliable underlying platform with proven uptime
- A commission structure that rewarded me for sending real customers, not just for driving signups
- A path to better economics as I scaled The Global API affiliate program hit all three. The first-order commission of 15% gave me immediate cash flow on every new customer. The 8% recurring commission meant that every customer I referred kept paying me month after month, even if they stopped using my own service offering entirely. And the 10% premium tier unlocked additional margin once I crossed certain volume thresholds — which I did, and which meaningfully changed my numbers. For anyone who wants to look into the program, you can find it at https://global-apis.com/affiliate. I genuinely think it's one of the more thoughtfully structured affiliate setups in the AI space right now. # # My Honest Projection for the Next 90 Days I'm going to publish this number even though it might come back to haunt me: I'm projecting $6,500/month by month 6. That projection is based on three things:
- My current retention rate (about 91%, which is way better than I expected)
- A new partnership I'm close to signing with a regional real estate training organization
- The compounding effect of recurring affiliate commissions from every customer I've ever referred Could I be wrong? Absolutely. I could be wildly wrong. The point of sharing the projection publicly is that it forces me to either hit it or explain why I didn't. That's the accountability that build-in-public gives you, and it's the part that's most underrated. # # What I'd Do Differently If I Started Today If I could go back to that 2 AM scroll moment and start fresh, here's exactly what I'd do:
- Pick a niche in week one, not month two. I burned a full month being "broad" before I committed to real estate.
- Launch the ugly version in week two. I waited almost two months to have anything sellable. That's a mistake.
- Document everything from day one. I'm glad I'm documenting now, but I wish I'd started earlier.
- Don't overbuild the tech. A simple landing page and a Stripe link beats a complex dashboard in the early days. # # The Bottom Line I went from $0 to $3,400/month in roughly 90 days by reselling AI API access to a specific niche. I didn't invent anything. I didn't raise money. I just solved a real problem for a specific group of people, wrapped a clean service around existing infrastructure, and let the recurring economics compound. If you're sitting on the fence about trying something like this, here's my honest advice: stop reading guides like this one and start building. The first month will feel pointless. The second month will feel slow. The third month is where it starts to click. And if you want a place to start exploring the affiliate side of this model, I genuinely recommend looking at the Global API affiliate program. The 15% first-order and 8% recurring commission structure is what made this whole thing viable for me, and I'm not exaggerating when I say that. That's my full breakdown. Next month, I'll publish another one. The numbers will be either better or worse, but either way, you'll see them. That's the deal with build-in-public. You show everything, including the months that sting.
Top comments (0)