Every Chartered Accountant working with Tally faces the same problem: clients hand over bank statements as PDFs, but Tally doesn't import PDFs. Something has to bridge that gap — and for most firms, that something is still manual data entry.
This guide covers every method for converting a bank statement PDF to Tally-compatible XML, what each approach costs in time and money, and why AI conversion has become the default for high-volume practices.
Why Tally needs XML, not PDF
Tally Prime and Tally ERP support a specific XML format called TallyXML — a structured data format that tells Tally exactly how to create vouchers. When you import a TallyXML file, Tally reads each entry and creates Payment or Receipt vouchers automatically, with the correct date, narration, and amount.
PDFs, on the other hand, are essentially images of documents. There's no structure Tally can read — each bank lays out its statement differently, uses different column positions, and formats amounts differently. Converting a PDF to TallyXML requires understanding that layout and extracting the underlying data.
Method 1: Manual data entry (the old way)
Most firms still do this. Someone opens the PDF on one monitor and Tally on another, and types each transaction one by one. For a 300-transaction statement, this takes:
- 3–5 hours for an experienced accountant
- 5–8 hours for a junior staff member still learning Tally
- An error rate of approximately 1–3% (1 mistake every 33–100 entries)
The errors are the bigger problem. A missed decimal point or wrong date can cause a bank reconciliation to fail and take another hour to trace. At scale, manual entry is genuinely expensive — not just slow.
Method 2: Copy-paste from bank's Excel export
Some banks offer a statement download in Excel or CSV format. You can use these formats in combination with a macro or a Tally import tool to create XML. The process:
- Download statement as Excel/CSV from the bank portal
- Clean and reformat the data (remove headers, fix date formats, separate debit/credit columns)
- Use a Tally XML template to build the import file
- Import the XML into Tally
This works — but the "clean and reformat" step is deceptively hard. Each bank formats its Excel export differently. HDFC uses one date format; SBI uses another. Some banks put debits and credits in separate columns; others use a single "Amount" column with a +/- sign. Every new bank means a new cleanup recipe.
It's also not an option when you only have a PDF (which is most of the time — many clients just scan or photograph their statements).
Method 3: AI-powered PDF conversion (Doxify)
This is how the workflow looks with an AI conversion tool:
- Upload the PDF. Go to doxify.in, sign in, and upload the bank statement PDF. Any PDF from any bank works — no templates, no setup.
- AI reads the statement. Doxify uses Claude AI to read the statement the way a human would — identifying the date column, narration, debit, and credit regardless of how the bank has laid out the page.
- Review the preview. Before downloading anything, you see a table of every extracted transaction. You can edit any line — correct a misread narration, fix a date, split a combined entry.
- Download the TallyXML. Click Download and you get a properly formatted XML file ready for Tally Prime or Tally ERP.
- Import into Tally. Gateway of Tally → Import → Vouchers. Select the file. Done.
Total time for a 300-transaction, 12-page statement: under 2 minutes.
What does the Tally XML contain?
Each transaction in the XML becomes a Tally voucher with:
- Voucher type: Payment (for debits) or Receipt (for credits)
- Date: Extracted from the statement
- Narration: The full transaction description from the bank statement
- Amount: Debit or credit amount
- Reference number: Cheque number, UTR, or transaction ID where available
Tally then places each voucher into the correct date in the Day Book, ready for bank reconciliation.
Which banks does AI conversion support?
Because AI reads bank statements like a human rather than pattern-matching against templates, it works with any bank. Doxify is tested against all major Indian banks:
- HDFC Bank, State Bank of India, ICICI Bank, Axis Bank, Kotak Mahindra Bank
- Union Bank of India, Bank of Baroda, Canara Bank, Punjab National Bank
- YES Bank, IndusInd Bank, IDBI Bank, Federal Bank, South Indian Bank
- International banks: DBS, Standard Chartered, Citibank, HSBC
New bank formats are automatically handled — there's no template update needed when a bank changes its statement layout.
Accuracy: how reliable is AI extraction?
Doxify achieves over 99% accuracy on standard bank statement formats. The main sources of error are:
- Scanned PDFs: If the PDF is a photograph or scan rather than a native digital PDF, OCR quality affects accuracy. Digital PDFs (downloaded directly from the bank portal) are always more reliable than scanned copies.
- Page breaks mid-transaction: Some banks split a long narration across two pages. Doxify handles this with a two-pass parser, but very unusual formats may require a manual review.
- Mixed-currency statements: Statements with transactions in multiple currencies require extra attention during the review step.
The preview step before download exists precisely to catch these edge cases. You review, edit if needed, and then download — so the final XML is always clean before it enters Tally.
How is pricing structured?
Doxify prices by PDF page, not by conversion. Plans start at ₹99 per year for 100 pages. A typical CA handling 20 clients a month, each with one 10-page statement, would use 200 pages a month — covered by the Standard plan at ₹599/year (900 pages).
Every new account also gets 100 free pages to test with real client statements — no credit card needed.
Is client data secure?
Doxify processes uploaded PDFs entirely in memory. No transaction data is written to disk or stored after the conversion is complete. Files exist only during the conversion window and are discarded immediately after. Client financial data never persists on Doxify's servers.
When should you still use manual entry?
Manual entry still makes sense in a few specific cases:
- A client has 5–10 transactions per month (faster to enter manually than to log in and upload)
- The bank statement is a physical printout with no digital version available (scanned PDFs still work, but with lower accuracy)
- The statement is in a very unusual format not yet well-handled by AI
For everything else — and that's the vast majority of real-world CA work — AI conversion is faster, cheaper per transaction, and more accurate than manual entry.