People ask me if AI is actually making anyone money, or if it’s all chatbots and hype. Here’s my honest answer, from my own business.
I’m a 67-year-old retired guy, and I run a small Amazon FBA business. Just me. No employees, no office, no team. I’m not going to tell you what I sell, because that’s not the point, and frankly it doesn’t matter. What matters is this: my business now gets work done every day that, not long ago, would have needed a room full of people or a stack of expensive software subscriptions. Most small sellers can’t afford either. I don’t have either. I have AI, and I have the right partner on the warehouse floor. Between the two of them, I spend about three hours a week on the business.
If you haven’t heard the term, FBA stands for Fulfillment by Amazon. You send your inventory to Amazon’s warehouses, and Amazon stores it, packs it, ships it to the customer, and handles returns. It’s how a one-person seller can offer Prime shipping to millions of buyers. But FBA only handles the last mile. Everything before that, what to buy, what to charge, and when to send more, is still on you. That’s where AI comes in.
This isn’t a how-to. It’s a look at how far one person can go right now, what I’ve actually built, and what I’m still working on.
Everything used to be done by hand
When I started, every part of this business was manual. I’d read a supplier’s catalog line by line. I’d look up each item’s sales history. I’d guess at what to buy and how many. I’d set prices and then forget to change them. I’d check stock levels when I remembered to, and build shipments in a spreadsheet.
Today, almost all of that runs on its own. On a normal day my system runs 79 different scheduled jobs, about 1,950 times. Every night it checks itself with more than 940 automated tests, and it tells me if anything is wrong before I’d ever notice.
Running my Amazon FBA business used to take me a full week of effort, and then some. Today it takes me about three hours a week. That turned out to be the biggest surprise of all, and I’ll come back to it.
I still make the calls that matter. I decide what to buy, and I approve every order that goes out. But I’m making those calls with information I never used to have, and I’m not spending my days doing the grunt work.
Buying smarter
Buying is where a business like mine lives or dies. Buy the wrong thing and your money sits on a shelf.
Now, when a supplier sends me a catalog, the AI reads every line. It pulls the sales history for each item, how prices have moved over time, how many other sellers are competing, and what every Amazon fee would be. Then it works out the most I can pay for each item and still make the return I want. Items get sorted into buy, make an offer, or skip.
And it doesn’t stop at the purchase. Every buy gets graded afterward. Did that item sell the way the analysis said it would? If it didn’t, that’s a flaw in the analysis, and the analysis gets fixed. That feedback loop is something I could never have done by hand.
Pricing with a gut feeling, not a rulebook
Most Amazon FBA sellers price with repricing software. You set rules like “beat the lowest price by a penny” or “never go above this number,” and the software follows them blindly, all day, for every product the same way. I don’t do that.
Instead, my AI and I worked together, product by product, on how I thought each one should move strategically. I’ve been doing this a long time, and I have opinions and painfully learned experiences. When should a product hold its price because buyers will pay it? When is it worth stepping down to get a sale moving? When should you sit tight and let the competition sell out first? I didn’t hand the system a list of rules. I taught it how I think about those calls.
Now it has its own trained gut feeling about the right move. It reads each product’s situation the way I would: how it’s been selling, who’s competing, what buyers have actually paid, and where the market is heading. Then it makes a judgment call. Not a formula. A gut feeling, built from mine. It does that for around 440 products, around the clock, which is something I could never do by hand.
It does have a couple of hard safety lines. It will never sell below what a product cost me, for example. But inside those lines, the decisions are judgment, not rules.
And I don’t just trust that gut. I check it. A lot of the “everyone knows” advice about selling on Amazon turns out to be wrong when you test it against your own sales, and a gut can be wrong too. So right now seven controlled experiments are running, where some products follow one approach and a comparison group follows another, so I can see which one actually makes more money. When the results come in, the system’s gut gets better. That’s the kind of thing big retailers have data science teams for. I have one AI and a laptop.
Restocking just in time
This is the part that would have been unthinkable for me a year ago.
My system watches stock levels in Amazon’s FBA warehouses against how fast each item sells. When something is about to run low, it builds the restock list, works out how many units to send, checks that Amazon will accept them, creates the shipping plan, and sends the order straight into my fulfillment partner’s system. No email. No spreadsheet. No phone call.
For the holiday season I’ve moved from one big weekly truck shipment to small orders on any weekday, so popular items get back in stock fast. When it makes sense, I still send bigger weekly shipments by freight truck, which costs less per unit but takes longer to be checked in. The system tracks how long each one takes so I can pick the right option instead of guessing.
Then, when the order ships, the shipment details flow back on their own, usually within the hour. This morning I got an email that an order was complete. By the time I asked about it, my system had already recorded the shipment, matched it to the plan, and logged it as on its way to Amazon.
Why the right fulfillment partner is everything
Here’s the part people underestimate. All the AI in the world doesn’t help you if the physical side of the business can’t keep up.
Carolina Prep & Ship is my fulfillment partner. They receive everything I buy, count it, store it, manage the inventory, prepare it to Amazon’s standards, and ship it out, to Amazon and to my other sales channels. They’re the yin to my yang. I handle the decisions. They handle the physical world. And it only works because both halves talk to each other.
For the record: CPS isn’t paying me anything to say this, and nobody asked me to write it. I’m a paying customer. I’m bragging about them because together we’ve been an awesome team.
I asked Matt Weschler, who owns CPS, why they built their own system instead of running on email and spreadsheets like a lot of warehouses do.
“Because email and spreadsheets lose things. A count lives in one person’s inbox, a shipment lives in another’s, and nobody sees the whole picture. … If Steve’s system can read our numbers directly and send orders straight to our floor, nobody retypes anything, and retyping is where most mistakes come from.”
That system is called Mission Control, and they opened it up so my system can talk to theirs directly. It can see their live inventory of my products. It can send orders. It gets notified when goods are received and when shipments leave. Their system and mine are, in a real sense, working together.
I was curious what it’s like on his end, having a one-person client whose orders come from an AI.
“Honestly, it holds us to a higher standard, and that’s a good thing. Orders show up complete and already checked. … What surprised me was how closely it watches. If a count is off by two units, it notices, and it asks. Early on it found problems that were ours, not his, like products we had received without the right identifier on file. We fixed them and told him so.”
Two systems that get better every day
This is the part I’m proudest of. The two systems don’t just pass data back and forth. They make each other better.
When my system notices that a number on my side doesn’t match a number on theirs, it flags it, and we sort it out together. When I needed shipping paperwork to come through the system instead of by email, their team built it the same day. When I wanted to route other kinds of orders through it, they built that too. I send them ideas. They send me ideas. Every week the whole thing gets a little tighter.
And some of what we fixed together now helps everybody they work with. Matt again:
“And the identifier problems his system caught changed how we receive: API clients now need a barcode on file at receiving, and every customer’s record flags stock with no identifier, so we catch it instead of the customer. That one helps every client, not just Steve.”
That’s not a vendor relationship. That’s a partnership. And I don’t think a one-person business could do what mine does without it.
How fast this has moved
What still amazes me is the speed. The system I vibe coded and run on today has had more than 8,800 changes since January. Things that were “someday” ideas in the spring are running every day now. Linking my system directly to CPS went from a request in September to live orders a few weeks later.
Here’s a number that still stops me. The software behind my business is now more than 500,000 lines of code, across nearly 1,900 files. I started it about nine months ago. That works out to roughly 2,000 lines of working code a day, every day, including weekends. No one person could write that by hand at that pace. The AI wrote the code. I decided what it needed to do, and checked that it did it.
And almost half of those lines aren’t the business logic at all. They’re tests: code whose only job is to check that the rest of the code is doing what it’s supposed to. That’s the part that lets me trust it with my money.
Before our system, I paid up to $700 a month for different software packages. I don’t pay that anymore. And that doesn’t even count the custom programming I used to pay for. In my life, I’ve paid over $100,000 for custom programs that specifically met my business needs. But here’s the thing: because of AI, I did all of this myself. It’s a great time to do something like this with AI.
What I am still working on
Making my buying even more accurate, so every purchase is graded against what was predicted.
Sharpening the pricing gut by testing its calls against real sales results.
Shortening the time from a restock decision to the product being sellable on Amazon.
Tightening how my system and CPS’s system stay in sync, so neither side ever has to chase the other.
The surprise benefit: I got my week back
Here’s the part I didn’t see coming. I set out to make my business smarter: better buys, better prices, fewer empty shelves. I wasn’t trying to get my time back.
But that’s what happened. Running my Amazon FBA business used to take a full week of my time, plus outside employee hours, specialty software packages, and more. Today it takes just me about three hours a week. That’s more than 90% of my working time back, while the business does more than it ever did when I did it all by hand.
I didn’t build this to step back from my own business. It just turned out that way. And honestly, that’s become the best part.
Would I sell it to you? No.
People ask if I’d sell access to the system I’ve built. The answer is no.
Frankly, I don’t want to provide support for other people’s issues. That might be a bit selfish, but it’s honest. This system was built around how I think and how I run my business. Turning it into a product would mean turning myself back into a full-time employee, just of a different business.
And I’m enjoying having my week back too much to give it away again. I’m going to enjoy the quality of life. Or maybe I’ll go start the next business.
But if you begin now, study how to vibe code, practice, fail, and learn, then you can do this too.
The honest part
AI isn’t magic, and it makes mistakes. It can be confidently wrong. So I built a lot of checks around it, and I still approve the decisions that spend real money. AI didn’t replace my judgment. It gave my judgment better information and took the busywork off my plate.
Matt put the limit better than I can. His team uses AI to build and run their own software, and he still says:
“But people still matter most on the floor. Somebody has to open the box, see the crushed corner, count the units and pack it right. AI can tell you a count is wrong. It can’t go find the missing case.”
If you’re running a small business, here’s what I’d tell you: the tools are here, they’re getting better by the month, and they’re within reach. But pick your partners carefully. The best AI in the world can’t fix a warehouse that loses your stuff. I found one that’s as serious about getting it right as I am.
And if you think you’re too old, or not technical enough, for any of this: I’m 67 and retired. If I can do this, a lot of other people can too.
You’ll notice I haven’t told you how much I sell. That’s personal, and I’m keeping it that way. But for one retired guy working about three hours a week, I’m proud of what I’ve built.
Matt Weschler owns Carolina Prep & Ship in Apex, North Carolina (FBA and Walmart prep, order fulfillment, receiving and storage): carolinaprepship.com. CPS didn’t pay for this post and didn’t ask for it. I’m a paying customer.