Using AI for business doesn't have to mean an enterprise contract and a six-figure invoice. One person who'd just started a small tech repair shop needed a way to track customers, repairs and parts — the exact job that usually comes with a monthly software bill. Instead of shopping around, he asked Claude, Anthropic's AI assistant, to write the app for him. He ended up with two working programs, a bill that's $300+ a year lighter, and a slightly stunned post online saying he gained nothing from the share — no ads, no upsell, just open source code anyone can download.
What happened
He was hunting for a ticket system for tracking customers and repairs. What he found was the usual wall: per-seat pricing, modules he didn't need, and a subscription for the privilege of using them. So he tried a different route — he asked Claude to build the thing instead.
In plain terms, a "ticket system" is just the numbered claim ticket your dry cleaner staples to your shirt. Every repair gets a number, a customer, a status and a history. Normally you rent that from a software company forever. Here it got written from scratch.
Two projects, not one
The first was a planning and customer-ticket app with inventory tracking — the day-to-day backbone of the shop. Then, in his words, he "got carried away" and built an all-in-one diagnostic tool for the repair bench as well.
The organizer he shared is a single Windows app that bundles notes, a calendar, customer records, tickets, money tracking and a locked vault. Your data stays on your own computer instead of sitting in somebody else's cloud. It's open source, free to download, free to edit, and there's no monetization at all.
The $300+ detail
He estimates he's saving upwards of $300 a year on paid software. That's the quiet part of the story: $300 isn't life-changing money, but it's recurring money, and what it bought was features that exist only because he asked for them. He already uses Zoho Mail and QuickBooks, so he wired both into his own app through their APIs — developer doorways that let two programs talk to each other. No plugin marketplace, no upgrade tier, no support ticket.
What it means for you
The interesting bit isn't the repair shop. It's that the gap between "I wish this existed" and "this exists on my laptop" has collapsed. Software is now something you can describe in a sentence instead of buy in a store.
At home
A shared shopping list, a household budget that doesn't fight you, a chore rota the whole family can see. You don't need to be technical — you need to be specific. Talk to it like you'd brief a helpful intern on their first day.
At work
Every team has one ugly spreadsheet held together with hope. A status tracker, a shift rota, a simple form that spits out a summary — those are exactly the little tools nobody budgets for and everybody needs. Describe the columns, describe what should happen when a row changes, and let the assistant draft it.
In your own business
The repair shop is the template. Tickets, inventory, invoices, follow-ups. If you already pay for Zoho Mail or QuickBooks, an AI-written layer on top can pull those numbers together without you buying a fourth subscription. Same idea, your trade.
For studying
Flashcard drills, a reading-log tracker, a simple quiz generator for a topic you're revising. Students are the ideal users here: the requirements are small, the answers are checkable, and the budget is usually zero.
For creative projects
A shot list, a clip library with tags, a submission tracker for the writing you keep sending out. Creative people are constantly improvising tools anyway — this just removes the part where you fight with formulas.
As a side income
He's giving his away, but the skill itself is sellable. Plenty of small local businesses need a tiny custom tool and can't justify a developer's day rate. If you can describe the problem clearly and test the result, you can charge for that.
How to try it right now
The tool to name here is Claude. The free option comes first and it's genuinely enough to start.
Step 1 — Open Claude on its free plan
Go to Claude in a browser and stay on the free tier. You'll hit usage limits on long sessions, so plan to work in a few chunks rather than one marathon.
Step 2 — Write your spec in plain English
Don't write code. Write what you want, the way you'd explain it to a colleague: "I fix laptops. I need to log each customer, what they brought in, what's wrong with it, what I quoted, and whether it's done." If you'd rather start from prompts other people have already tested, the free prompt library at MyKreaTool is a reasonable shortcut.
Step 3 — Ask for one small piece at a time
This is the single biggest factor in whether it works. Ask for the customer list first. Get that running. Then ask for the repair tickets, then the money tracking. Ten small requests beat one enormous one every time.
Step 4 — Ask for something you can actually run
Ask for a single self-contained file — a simple web page that saves data in your browser, or a script you can double-click. That way there's nothing to install and nothing to break.
Step 5 — Test it with fake data first
Type in three invented customers before you put a single real phone number in it. You'll find the rough edges immediately, and you can paste the error message straight back into the chat.
Step 6 — Add the integrations last
Once the basics work, ask about connecting your email or accounting software through their APIs — that's what the repair shop owner did with Zoho Mail and QuickBooks. It's the last step, not the first.
Step 7 — Or skip the building entirely
If you just want the finished thing, his Windows organizer is free and open source, with notes, calendar, customers, tickets, money and a locked vault in one app, and everything stored on your own machine: the organizer he shared.
Upsides and what changes
The obvious win is money: $300+ a year back in your pocket, recurring, without a cancellation phone call. The second win is fit. Off-the-shelf software is built for the average customer, which means it's slightly wrong for everyone. When you spec it yourself, you get the two features you actually use and none of the twenty you don't.
The third win is ownership. Open source and local storage mean the tool doesn't vanish when a vendor pivots or raises prices. Nobody can deprecate your own app.
And there's a mindset shift that's harder to price. Once you've watched a working tool appear from a plain-English description, you stop accepting bad software as weather. You start thinking, "I could just ask for that."
Limitations
Be clear-eyed about this. AI-written code can have bugs, and there's no support line when something breaks — you're the IT department now, or at least the person who pastes the error back into the chat. A "locked vault" feature inside a homemade app is only as private as the code behind it, so keep backups and don't treat it as bank-grade security. Integrations with services like Zoho Mail and QuickBooks can break when those services change their APIs, and you'll be the one fixing it. Free plans have usage limits, so bigger projects take patience. And if you can't describe what you want precisely, you'll get something confidently wrong — the tool is only as good as the spec you hand it.
Conclusion and one action for today
A repair shop owner wanted a ticket system, found the prices annoying, and asked Claude to write one instead. He saved $300+ a year, got two apps shaped exactly to his workflow, and gave them away for free. Nothing about that story required a computer science degree — just a clear description and a willingness to iterate.
Your one action for today: open Claude on the free plan and write five sentences describing the small tool you keep wishing you had. Don't ask for code yet. Just write the spec — the customer, the fields, the one thing you want it to do. That sentence is the whole starting point.
Comments 0