How I rebuilt ten years of accounts in a single day

For ten years, Oliverlist's finances lived in spreadsheets: tabs, formulas and, over time, 28,719 transactions. This year, I replaced them with a finance tracker connected to my three banks. It took one day.

If you run a company, the picture may sound familiar. Several banks, sometimes several currencies. Monthly figures that arrive from the accountant weeks after month-end. A spreadsheet that only one person knows how to update. The tool matters less than the method, and the method works in any jurisdiction. Here is how I did it, what the tracker found, what changes every week, and how to do the same in your company.

Ten years of spreadsheets, one day of work

The starting point: 28,719 transactions, ten years of history, three banks. Each line had a date, an amount, a bank description that was more or less readable, and a category assigned by my formulas.

I built the tool myself, with Claude Code, in one day. The logic fits in three steps.

1. The whole history in one place. All 28,719 transactions went into a single database, one row per transaction.
2. Written classification rules. What my formulas did silently became a list of readable rules: this description, that category. When a transaction matches no rule, the system asks me, and my answer becomes a new rule.
3. One reader per bank. Every bank formats its statements its own way. A small program for each bank reads them, so new transactions come in without anyone retyping them.

From that single table, everything else is calculated: the P&L, the forecast and the 60-day cash position. Every week, an agent adds the new statements and refreshes the whole picture.

As for running costs, the 84 agents behind my work as CEO run on a small server and a database, about $59 a month on top of the AI subscription. The finance spreadsheet has been retired.

What the tracker found

Moving ten years of transactions into one table means going through every line again. Two issues came out.

368 duplicates. Transactions recorded twice. A single duplicate is invisible in a summary table. Hundreds of them distort the totals, and every comparison built on those totals.

A VAT discrepancy. I will not share the amount here. What matters is how it surfaced: by classifying every transaction, then comparing the VAT actually paid with the VAT that should have been paid. Nobody runs that comparison by hand across ten years of lines.

I had not spotted either of them. It is not a matter of attention but of volume: beyond a few thousand lines, you only look at totals. With VAT and corporate tax filings, that blind spot is hard to live with.

What changes every week

The figures already existed. What changes is how long they take and how much effort they cost. Every week, with no data entry, I see:

• the P&L, by category, comparable month on month and year on year;
• the forecast, updated every week;
• the 60-day cash position, which shows a dip while there is still time to act;
• overdue invoices, with reminders drafted as soon as a payment is 7 days late.

That last point matters more than it seems. A reminder drafted on day 7 only needs my approval. One that waits for someone to remember it rarely goes out on time. I no longer write these reminders: I read them and approve them.

For a CEO, the difference shows in decisions. A hire, an investment, longer payment terms for a key client: you decide on this week's figures, not on last year's statements. Meetings with your accountant can then focus on variances rather than on getting the numbers straight.

How to do it in your company

You do not need my tool. You need the same method, applied to your data.

1. Gather the history. Bank exports, spreadsheets, extracts from your accounting software. Do not clean anything first: cleaning is part of the job.
2. Build one transaction table. Date, amount, currency, bank, description, category. The P&L, cash position and forecast are calculated from it, never alongside it.
3. Write down your classification rules. They already exist, in your formulas or in the head of whoever keeps your books. Once written, they can be read, corrected and applied automatically to every new line.
4. Check before you trust. Calculated balances must match your bank statements, bank by bank, month by month. Look for duplicates. Until the balances match, do not steer anything with it.
5. Automate the update. An agent adds new statements every week, recalculates everything and flags what it could not classify.
6. Keep your accountant in the loop. This tracker is for running the business. It does not replace your bookkeeping or your tax filings.

How long will it take in your company? I cannot say without seeing your files. For me, one day. For you, it depends on the number of banks, entities and currencies, and on the state of your history: that is what a two-week audit measures. A finance tracker also makes a good first sprint: one tool in production in 14 days, used on your real workflows. Either way, accounts and subscriptions stay in your name, and everything is documented on your side.

Start with 30 minutes

If your monthly figures arrive weeks after month-end, or your cash position depends on a file only one person can update, this is probably where to start.

Book 30 minutes: you walk me through how your figures are kept today, and I tell you what can run without you. If you would rather start on your own, take the online test: 10 questions, your level among four, and the 3 tasks to hand to AI first. The result appears on screen, and nothing is sent or stored.