Statemently converts bank statement PDFs to spreadsheets - but unlike OCR converters, it reconciles every debit and credit against opening and closing balances and flags any uncertain row for review. Exports to Excel, CSV, QuickBooks (QBO), OFX, QIF, and Google Sheets. Every major US bank (Chase, BofA, Wells Fargo). Processed in memory, never written to disk, auto-deleted after conversion - no account required to try. Pricing: one-off credits or Pro at $49/mo (~100 statements/mo, all banks).
No reviews yetBe the first to leave a review for Statemently
Maker
π
Hey PH, Tamas here, solo maker. Statemently came from watching bookkeepers babysit bank-statement PDF converters - the OCR ones drop or misread rows and you don't find out until the books don't balance. So Statemently reconciles every debit and credit against the statement's opening and closing balances, and flags any row it's not sure about instead of silently guessing. Exports to Excel, CSV, QuickBooks, OFX, QIF, Google Sheets. It runs in memory, nothing is written to disk, and files auto-delete after conversion - no account needed to try. Two things I'd love feedback on: (1) which bank/format trips up your current converter the most? (2) does balance-reconciliation + flagging uncertain rows earn the trust that plain OCR doesn't? Thanks for looking.
Report
How does the reconciliation actually work when a PDF statement has scanned or misaligned numbers - do you use a specific parser per bank, or is it more of a generic rules engine that learns over time?
Report
Maker
@hasantlicΒ Great question β it's neither. No per-bank parsers, no learning rules engine.
Extraction is handled by a vision-capable LLM that reads the statement layout directly, so it generalizes across banks without templates and holds up on scanned or misaligned figures where coordinate-based parsing tends to fail.
Reconciliation is a separate, deterministic check: we sum every signed debit and credit and confirm the total equals the statement's own closing-minus-opening balance, to the cent. If it doesn't tie out, we flag the exact discrepancy rather than return numbers we can't stand behind.
The AI reads; the arithmetic verifies. Accuracy doesn't depend on the model being perfect β it depends on the math balancing.
Report
The balance reconciliation is a nice touch - so many PDF converters miss that and you end up with a spreadsheet that doesn't actually add up. Quick test on a Chase statement came out clean.
Report
Maker
@edanurbatu11295Β Thank you β that quietly-doesn't-add-up spreadsheet is exactly the failure mode we set out to kill, so I'm glad the Chase statement came through clean. If you're on QuickBooks, you can now push the verified rows straight in. And if you ever hit one that won't reconcile, send it our way β those are the cases we most want to see.
How does the reconciliation actually work when a PDF statement has scanned or misaligned numbers - do you use a specific parser per bank, or is it more of a generic rules engine that learns over time?
@hasantlicΒ Great question β it's neither. No per-bank parsers, no learning rules engine.
Extraction is handled by a vision-capable LLM that reads the statement layout directly, so it generalizes
across banks without templates and holds up on scanned or misaligned figures where coordinate-based parsing
tends to fail.
Reconciliation is a separate, deterministic check: we sum every signed debit and credit and confirm the total
equals the statement's own closing-minus-opening balance, to the cent. If it doesn't tie out, we flag the
exact discrepancy rather than return numbers we can't stand behind.
The AI reads; the arithmetic verifies. Accuracy doesn't depend on the model being perfect β it depends on the
math balancing.
The balance reconciliation is a nice touch - so many PDF converters miss that and you end up with a spreadsheet that doesn't actually add up. Quick test on a Chase statement came out clean.
@edanurbatu11295Β Thank you β that quietly-doesn't-add-up spreadsheet is exactly the failure mode we set out to kill, so I'm glad the Chase statement came through clean. If you're on QuickBooks, you can now push the verified rows straight in. And if you ever hit one that won't reconcile, send it our way β those are the cases we most want to see.