Every bank, a different page
Ask two banks for a statement and you get two documents that share almost nothing — different column orders, different date formats, running balances in one and not the other, footnotes and marketing wedged between the transactions that matter.
The old tools answered this by asking you to build a template per bank, then another when the bank redesigned its PDF. It is a treadmill: the format you taught the machine yesterday is not the format that lands in the inbox next quarter.
From pages to rows
Docusift reads the statement the way a person would — understanding what a transaction is, rather than matching a saved layout — and turns each line into a structured row: date, description, amount, balance, whatever the page carries. No template for this bank, no training run, no schema file to maintain.
Every field comes back with a confidence score, so a smudged amount or an ambiguous date is flagged rather than quietly guessed. The clean rows export to Google Sheets or push into the ledger you keep, ready to reconcile instead of ready to retype.
Reconciled, not just read
Reading a statement is only half the job; the point is rows you can trust enough to match against what you already have. Because the low-confidence fields surface for a human before anything lands, reconciliation starts from checked data rather than from a fresh pile of copy-paste errors.
Docusift is live and accepting sign-ups now, and the statements you upload are yours — processed, returned, never resold. The same reader handles the invoices and receipts sitting next to the statement in the same inbox.