Statement Categorizer

The problem

Imran’s firm, Global Tax Corporation, sorted client bank statements into expense categories by hand. A spreadsheet of keywords caught about three transactions in four. Someone read the rest line by line.

What I built

A tool the firm runs on its own computer. Staff drop in a statement export from any Canadian bank. It pulls out every transaction and sorts the expenses: first by the firm’s own rules, then by matching things it has seen before, and only then by asking AI. Deposits are set aside automatically. Results show within minutes of importing.

The accountant checks everything on one screen, grouped by business, with the biggest amounts first. Each answer becomes a rule for next time, so the tool needs the AI less every month. A person always makes the final call.

Only cleaned-up business names and amounts are sent to the AI. Never client names, account numbers, dates or balances. The firm can see what the AI cost on every statement, about ten cents for a full year.

Where it is now

Imran’s team uses it heavily on real client statements. Once its rules had built up, about 90 percent of the transactions on real statements sorted themselves.

Available to other firms

Any firm can run it with its own rules and categories. If you sort statements by hand, ask me about it.