Stop Rebuilding Your Books Every Year

A Saturday in March. A year of bank and card statements open on the screen, a coffee going cold, and the same question on repeat, line after line: business or personal? Real deduction, or just money moving between my own accounts?

Here's what should bother you more than the lost Saturday: you answered most of these questions last March. The hardware-store run, the software subscription, the transfer to savings that is definitely not income — you already made those calls. Then they evaporated, and this year you're making them all again.

Sorting the statements was never the hard part. The hard part is re-making a year of judgment calls you already made once — because nobody wrote them down.

The map is the asset, not the spreadsheet

I found this out building the Statement Reconciler. Once a good categorization map existed — this account means this, that vendor goes there, transfers are never income — reconciling a whole year of statements shrank to a file drop, minutes of work. Every hour of pain lived in building that map, and it was being built backwards: run the statements, see what's wrong, fix one transaction, run it again.

That backwards order is the common failure, and it isn't a tooling problem. Roughly 95% of AI initiatives fail, and the ~5% that get results aren't using better tools — they follow a different sequence (AI READY, Danny de Vries). For your books, the sequence change is simple: decide the rules once, up front. Then let the software follow them.

Think of it like a tailor's measurement card. The first fitting is the slow one — someone actually measures you, shoulder to cuff, and writes it all down. After that, every garment fits without a tape measure coming out, because the card remembers your shape. Your books have a shape too: what counts as business, where each vendor lands, what's never income. The map is that card — your judgment, measured once and written down.

Taking the measurements takes an afternoon, not a season

The new free tool, Books Builder, drafts the card for you. Feed it last year's categorized books — a year of your own decisions is already sitting in there — and it mines out your real chart of accounts and vendor rules from how you categorized. No prior books? One fresh statement gives it enough to propose a sensible starting map.

Then an AI accountant pass reviews every account: business or personal, real deduction or just a transfer, and it flags the few judgment calls that actually move your taxes — ranked biggest dollars first. Your job shifts from making a thousand small calls to reviewing a short list of big ones. Accept or adjust each, save the map, and it exports straight into the Statement Reconciler — where next year's statements sort themselves.

Decide business-vs-personal once, in an afternoon, instead of a thousand times every March.

What it touches, plainly

Before you point any tool at money data, you deserve the data story in one breath. Your file is parsed in your browser and never uploaded. For the AI accountant review, only short text descriptions and per-account totals are sent — never amounts per line, never dates, never the file. The map you build lives in your browser only; export it to keep it. And everything the AI returns is a labeled draft — flagged suggestions you accept or adjust, nothing finalized until you say so. The AI proposes; you sign off. Your accountant still gets the final word.

Try it on your own history

Don't take the argument on faith — that's why the tool is free and ungated. Pull up last year's books, or one recent statement, and see what map comes back. I've shipped north of 20,000 software products, and the pattern that survives all of them is this one: the work you write down once stops being work.

You already made the decisions. This year, keep them.