1. A separate pipeline for statements
Bank statements go through a pipeline built specifically for transactions rather than line items: for each one, Pileform extracts the date, a counterparty hint, the amount, the direction (in or out), the running balance, and a type hint straight from the statement, in whatever format it arrives.
2. How a transaction gets classified
Classification runs in tiers, cheapest and most certain first: transaction-type defaults, then a fuzzy match against counterparties you've used before, then an AI fallback against your full chart of accounts, and finally a manual flag if none of those land with confidence. Transactions are grouped by counterparty, payee, account, or type, so a page of individual lines reads as a handful of coherent groups instead.
3. The output
The result is written to the workbook with proposed double-entry journal postings, mirroring how invoice and receipt data comes back, so the review experience is the same either way: see understanding the review screen for what happens next, or how review flags work for what triggers a manual check on a transaction.
Still stuck? Email contact@pileform.com and a person replies within one business day.