Converting the PDF was never the slow part — deciding which ledger each of 300 rows belongs to is. BankXL reads the narration your bank already printed and assigns a Tally ledger to every transaction, using 27 rules across 26 ledger heads.
It runs as plain text matching inside the same request that reads your PDF. No model, no external API, nothing about your client's statement leaves the server. You can read every rule that decided every posting.
Questions
Does this use AI to categorise transactions?
No. Categorisation is a table of text-matching rules that runs inside the same request that reads your PDF. There is no model and no external API call, so your clients' narrations are never sent anywhere.
Can I use my own ledger names from Tally?
Yes. Upload a two-column CSV — the text to match, and the ledger to post it to. Your rules are checked before the built-in ones, so your firm's nomenclature always wins. We store the rules, never the transactions they matched.
What happens to transactions that do not match any rule?
They are posted to Suspense A/c and left for you. We would rather hand you a blank than quietly file something in the wrong ledger, because a wrong posting is harder to find than a missing one.
Does the ledger appear in the Tally XML export?
Yes — it becomes the contra ledger on each voucher, so the XML imports into Tally already posted rather than landing in one large suspense pile.
Is it included in the free plan?
Yes. Categorisation runs on every conversion including the free tier. Uploading your firm's own mapping is part of the paid plans.