I am not a developer. The market research was done in my local pub. And last week the app went live on the App Store.
Yoley is a quoting, invoicing and payments app for UK tradespeople. It took six months, 155 releases since I moved to Claude, and about the price of a laptop in AI. It’s free, it was never meant to be a business, and it taught me more about product, about AI and about Monek than anything I’ve read in three years. Here’s the honest version.
The sum
I built it because of a sum I couldn’t get out of my head. Moore’s Law gave us fifty years of computing power doubling while the price fell. Put AI on top of that and the number of apps, tools and businesses being built is about to grow faster than at any point in history. Monek sits in the middle of thousands of businesses taking payments, and our technology — like everyone else’s in our market — was built for the payment, not for onboarding and engaging a world where choice and opportunity are multiplying. How do we serve that market at high quality and low cost? Which horse do we back? And how do we keep the service levels we’re so proud of when the number of things wanting to plug in goes vertical? The strategy we wrote three years ago is already completely out of date.
I couldn’t answer that by reading posts. I had to be the customer for once, and find out where Monek was deficient in this brave new world by trying to build on it myself.
The pub
So I went to the pub. It seemed as good a place as any to change the world, and I suspect many would agree. Tree surgeons, farmers, electricians. Every one of them described a world nothing like the one I thought I understood, for the price of a few pints. You cannot get that from a survey. The story of a tree surgeon is nothing like the story of a plasterer, and neither is anything like what an accountant thinks it is. An app has to understand the story before it can offer the solution.
Then I went home and asked Claude to run a full functional analysis of every competitor, reviews included, and in particular to find what people were unhappy about. That became the spec. Build what people already love. Solve what the competitors leave unanswered. All of it from the sofa, late at night.
Three versions
V1 was intoxicating. Your first run at building with AI comes so fast you feel like a genius. You aren’t. It’s slop. Poke-sticks-in-your-eyes ugly, poor flow, complicated navigation, far too many features. Don’t get high on your own supply of bullshit. I broke it, quickly and thoroughly, and that was the best thing that happened.
V2 started again, and the first thing I built wasn’t a feature. It was the factory: a full CI/CD framework on Claude, Xcode, Git and automated testing, built product-first rather than tech-first. Every feature ships with its tests and documentation generated alongside it, and the latest code standards are baked into every rebuild. When promoting a change carries no risk you stop hesitating, and fail forward becomes a method rather than a slogan. That’s how one person reaches 155 releases without leaving a mess behind them.
V3 was where it became a product. We stripped features out rather than adding them, until it stopped being a junkyard A-Team bodge. Cutting was harder than building and made more difference than anything I added.
If you know Simon Sinek’s what / how / why, you’ve just read my six months. V1 was pure “what” — a pile of features. V2 was “how” — the system that builds them. Only in V3 did I get to “why”: why does a tradesman open this at all, and what is he really trying to get done at 7pm sat in a van?
Sinek says start at why. He’s right, and AI has made him more right. It makes a developer five times more productive — which also means they can head in the wrong direction at five times the speed. “What” feels like progress and “why” feels like talking, so almost nobody starts there. Knowing why is mostly experience — every grey hair is a problem solved, bought and paid for. What AI changes is that the experience no longer has to stay with the person who paid for it. Put it in the spec, the tests and the factory and it flows forward, so the next person can build on it without growing the grey hairs themselves. The factory is what bought me the freedom to go back and ask the better question without it costing me the project.
The part nobody posts about
I’m a night owl: 9pm to 1am, every day, on top of a day job that was already demanding. Do not underestimate the work, or the discipline. The hardest moment was admitting this had stopped being a learning exercise. You don’t learn anything from half-built.
To come first, you first have to finish.
And the finished thing is what exposed Monek.
What it exposed
Integrating our own onboarding into a third-party app — as the customer, not the provider — showed me exactly where we fell short. We believed we were in the top 10% of onboarding experiences in this market. We probably were. But the gap between good and frictionless is a hundred small things you only see at 1am when you’re the one wiring it up.
So we rewrote the integration documentation, this time to be read by AI tools as well as people. Claude can now integrate Monek into an app or website in under three minutes, right first time. That used to be hours, sometimes days, for anyone meeting our docs for the first time. We revised the onboarding framework and built a way to embed account opening inside a partner’s app, so a merchant never leaves the product they’re trying to use.
That’s the answer to the sum. When the next ten thousand apps arrive, the cost of serving each one well is the cost of documentation an AI can read. None of it came from a strategy offsite. It came from failing at it myself.
The numbers
The costs, because people deserve honest numbers even when they’re round ones:
- A MacBook of its own, so nothing of Monek’s touched it
- £99 to Apple
- £29 to Google
- As much AI as your app needs — in my case, roughly the price of the laptop again
Three years ago this was a team of five, eighteen months and a budget north of £700k. And the budget was never the real cost. The real cost was the eighteen months: the revenue you don’t earn while you’re building, the customers who settle for something else in the meantime, and the competitor who launches in month nine while you’re still arguing about the roadmap. AI didn’t just make the build cheaper. It took most of the risk out of the calendar. What exists now is better built: the website, both apps, the testing framework and the documentation synchronise automatically to every change. No technical debt.
What changed
What I learned was product management and software design, not coding. Coding is increasingly for the bots. Developers are becoming architects and product managers, and the ones who make that move will be worth far more than they are today.
It may also change what a business is worth. For a long time the valuation conversation has started with IP. But when anyone can rebuild your app in six months for the price of a laptop, I’m no longer sure that’s where the value sits. Try Sinek’s frame on a valuation instead: is IP the “what”, product strategy the “how”, and loyalty the “why”? If it is, the thing worth paying for isn’t the code. It’s recurring revenue from a great product with customers who stay.
Which is why Yoley is free. It costs the tradesperson nothing; Yoley earns partnership income from Monek on the payments that flow through it, which means it only makes money while people keep using it. That’s the whole thesis in one line.
Full disclosure, since it matters: I’m the majority shareholder of Monek and I own the whole of Yoley. Yoley was built in a sealed box — my own laptop, my own money, my own evenings, and none of Monek’s people, code or systems. Partly because that’s the only honest way to replicate what the next generation of entrepreneurs will actually experience, and partly so the boundary is clean. Its only connection to Monek is the partnership: Yoley processes exclusively through Monek, on the same public integration any partner would use.
Yoley was built to do two things: educate me, and help Monek. Both done. What it became along the way is worth doing properly, so I’m recruiting a team to run it and to take the factory into other vertical markets. Tradespeople were the first; they won’t be the last. And I’m stepping away. The lessons come with me, back to Monek, which is where they were always meant to be spent. The Android version lands in a couple of weeks. I’ve told family and friends to prototype an idea a month and launch three or four a year — the first off that line is already out and looks superb. Well done James.
If you’re a trade running your books on paper, WhatsApp and hope: Yoley is free, with no monthly fee.
If you’re an executive still reading about AI rather than building with it: go to the pub, listen to someone whose job is nothing like yours, then go home and build the thing. And finish it. Then hand it on.