How I built an invoicing system with double-entry bookkeeping in 3 days

I needed an invoicing system. Not some generic SaaS, but exactly what I needed: double-entry bookkeeping, export formats for accountants, cash flow tracking, and bank integration.

Normally I’d have two options: buy an off-the-shelf solution that’s 80% right and 20% annoying. Or build it myself — and budget at least three weeks of focused work.

I chose a third path. Invoicing was done in one day. Double-entry bookkeeping in two more. Three days total.

What I built

This isn’t a toy. The system handles:

Under normal circumstances, this would take a senior developer at minimum three weeks. Possibly more, if they needed to get up to speed on double-entry bookkeeping rules.

The key: a specification against legislation

This is where it got different from other projects. Proper use of Claude Code didn’t just lead to faster code — it led to the preparation of a complete specification verified against legislation.

This would have taken one person weeks of manually studying laws and regulations. Claude Code handled it as part of the process — not as a separate step, but as the foundation everything was built on.

How it went

Day 1: Invoicing. The basic invoicing system — creating invoices, client management, due date tracking. This was the simpler part.

Days 2-3: Double-entry bookkeeping. This is where it got complex. Double-entry bookkeeping is a domain that normally requires consulting with an accountant. But thanks to the specification Claude Code prepared, I had a solid foundation.

And then the most interesting part: two iterations of the system were completed overnight. I gave Claude Code direction, went to sleep, and woke up to results in the morning. I reviewed, gave feedback, and the next night the same process repeated.

What actually changed

People think AI is faster copy-paste. It isn’t. This is a story about how AI can handle something a person alone couldn’t do in a reasonable timeframe: study the legislation, prepare a specification, and then build a system based on it.

The system is now going for review to an accounting firm. Not as a prototype — as a finished system, built on a specification verified against legislation.

What this means for you

If you have a project that requires deep knowledge of a regulated domain — accounting, law, healthcare — AI can dramatically shorten the path from zero to a working solution. Not by replacing the expert. But by preparing a foundation that the expert then validates.

I’m curious what that would look like in your context. Let’s talk.

You might also like

• A 7-week project in 1.5 hours
• How much does a day without AI cost you?
• The AI Pilot Trap: Why 77% of Enterprise Projects Never Reach Production