Digitize Accounting Documents: How to Automate with AI

Digitizing accounting documents comes down to three steps: scan the month's paper into one PDF, split that file into individual receipts and invoices with proper names, and upload the result to your accounting software or hand it to your bookkeeper. The middle step is where the hours go, and it's the one software can take over. Docusplit reads every page of the scan, finds where one document ends and the next begins, and names each file by its content: INV-2025-0105_Acme-Corp.pdf for an invoice, Receipt_Shell_2025-01-08.pdf for a fuel receipt. This guide walks through the full routine, including the pages the AI can't read.

Where the time goes

Scanning is fast once a document feeder does the work, and uploading files to QuickBooks or Xero takes minutes. The slow part sits in between: opening a 40-page scan, deciding where each document starts, cutting it apart, typing a name, moving the file. Done by hand, that's a minute or two per document, every week, for as long as the business exists.

The pattern looks the same in a bookkeeping firm whose clients drop off a shoe box each quarter, in a dental practice that scans supplier invoices at the front desk on Fridays, and for a freelancer who pulls the year's receipts together in January.

Worked example, with assumptions rather than measurements: 80 receipts and invoices a month at about a minute each for cutting and naming makes roughly 80 minutes a month, or 16 hours a year. In document mode the app budgets about 10 seconds per page, so the same 80 documents (say 100 pages) run through in 15 to 20 minutes, most of it waiting, plus a few minutes of spot-checking the result.

Step 1: Scan the whole period into one PDF

Paper documents. Load the entire stack into a scanner with an automatic document feeder and scan it as one PDF. 200 dpi works, 300 dpi is the safe choice; grayscale is fine for printed pages, color helps with receipts on tinted paper. Remove staples, flatten creases, keep the pages straight. There's no need to sort the stack first; sorting is what the AI does.

Digital documents. Save email invoices and portal downloads in one folder. You can upload each PDF on its own or merge a week's worth into one file; either way, each upload is one PDF of up to 50 MB with no limit on pages.

A weekly batch beats a monthly one. The piles stay small, a missing receipt is noticed while the purchase is still fresh, and a 30-page run finishes in a few minutes.

Step 2: Let the AI split and name the batch

Before uploading you make two choices. First the mode: document mode for a mixed pile of receipts, invoices, statements and notices, or invoice mode when the batch contains nothing but invoices (it reads fewer fields and runs at about 6 instead of 10 seconds per page). Then the folder layout of the ZIP: flat, by month (2025-01/) or by sender (Acme-Corp/).

Upload, watch the progress bar, download the ZIP. For a January batch in document mode with folders by month, the result looks like this:

Input:  scan_2025-01.pdf (20 pages, 8 documents)

Output:
2025-01/
├── INV-2025-0105_Acme-Corp.pdf
├── 0117-2025_City-Utilities.pdf
├── Receipt_Shell_2025-01-08.pdf
├── Receipt_Blue-Harbor-Cafe_2025-01-12.pdf
├── Statement_First-National-Bank_2025-01-31.pdf
├── Reminder_Northwind-Ltd_2025-01-20.pdf
└── Contract_Northwind-Ltd_2025-01-15.pdf
Unknown/
└── Document_Page_14.pdf
_overview.csv
_metadata.json

Two naming rules are at work. Pages the AI recognizes as invoices get InvoiceNumber_Company.pdf. Everything else gets Type_Sender_Date.pdf, where the type is a short label the AI writes from what it sees on the page (receipt, statement, reminder, contract, letter, notice) and the date is in ISO format so files sort chronologically. Page 14 in the example was a faded thermal receipt with nothing readable on it; it's kept as Document_Page_14.pdf in the Unknown/ folder rather than guessed. No templates or rules are set up beforehand. How the boundary detection works is covered in Automatic document separation.

Step 3: Upload to your accounting software

One document per file is exactly what the receipt-capture features in QuickBooks, Xero or Dext expect, so the ZIP contents can go straight in.

• Drag and drop: unzip, open the receipts or documents area of your accounting tool, drop the files in, and assign categories where the software asks for them.
• Bookkeeper hand-off: share the month's folder by cloud link together with _overview.csv, which lists every file with its number, sender and date and doubles as a checklist.

There is no API and no direct connector. The files are yours to move, which also means nothing gets posted anywhere without you looking at it first.

What typically goes wrong

The AI reads printed text well and struggles with the same things a tired human does. Handwritten receipts, very faint thermal paper, scans below roughly 200 dpi and blank back pages are the usual reasons a page comes back as Document_Page_N.pdf, with "Not detected" in the CSV and, if you chose folders by month, a place in Unknown/. A delivery note stacked behind its invoice can end up in the wrong file, an attachment without a letterhead may be glued to the wrong document, and a company name can come out slightly different from the one in your vendor list.

There is no correction step inside the tool. Check the CSV and the Unknown/ folder first, rename or move the handful of files that need it, and move on. For a 20-page batch that's usually a minute.

Keep the paper until you've checked the scan

Once the scan replaces the paper, retention rules apply to the file, and they differ by country: the IRS generally wants records for 3 years and up to 7 in some cases, HMRC 6 years for VAT records, Germany 8 years for invoices since 2025. Electronic storage is accepted in all three if the copies stay legible and findable. Keep the originals at least until you've confirmed that every page came through readable, and ask your accountant before shredding. The details, with sources, are in Invoice archiving and retention compliance.

Frequently asked questions

Can I destroy paper documents after scanning?

Often, but the rules are local. The IRS accepts electronically stored records if they are legible, indexed and the system is documented (Rev. Proc. 97-22). HMRC accepts a scanned image that contains all the VAT details. Germany expects a documented scanning procedure. Keep the paper at least until you have checked that every page is readable, and ask your accountant before shredding.

What scan resolution do I need?

200 dpi works, 300 dpi is the safe choice. Grayscale is fine for printed documents; use color for receipts with tinted paper or stamps. Below roughly 200 dpi, small print on receipts starts to fail.

How do I handle multi-page invoices?

The AI checks each page for signs of a continuation, such as "Page 2 of 3", a missing letterhead or a table that carries on from the previous page, and keeps those pages together in one file. Attachments without a letterhead are the usual exception.

How many documents can I process at once?

One PDF per upload, up to 50 MB, with no limit on pages per file. Only your page quota counts: 30 free pages once after creating an account, then 100 pages a month on Starter (€9.99), 300 on Pro (€29.99) or 1,000 on Max (€99.99).

How do I digitize documents that are already digital?

Email attachments and portal downloads go through the same process. Upload each PDF as it is, or merge a batch into one file first; the AI splits and names the result the same way as a scan.

Does this work with handwritten receipts?

Printed text is what the AI reads well. Clear block letters sometimes work; messy handwriting usually doesn't, and such a page is saved as Document_Page_7.pdf for you to name yourself.

Can I use this for tax season preparation?

Yes, and it works best as a weekly habit rather than a January marathon. Run each week's receipts through with folders by month, keep the CSV overviews, and your accountant receives one folder per month instead of a shoe box.

What file formats does the AI accept?

PDF only. Scanners and scanner apps export PDF by default; if you have JPG or PNG files, convert them to PDF first.

Start with last month's batch

Scan the stack, pick document mode, and see what the ZIP looks like: the first 30 pages are free after creating an account, no credit card needed. Start for free

Related articles:

• Digitize invoice processing: the practical guide
• Invoice archiving and retention compliance
• Auto-organize documents into folders