Auto-Sort Incoming Mail
AI classification and renaming
Last updated:
Scan the day's mail stack as one PDF and upload it. The AI splits it into individual letters and names each file by type, sender and date, like Notice_City-Utilities_2026-03-15.pdf. Bring your own scanner.
To split and rename a scanned mail stack, you upload the batch PDF. The AI recognizes from the content where each letter begins, keeps continuation pages together and names every file by type, sender and date, invoices by invoice number and company. You get a ZIP with one PDF per letter and a CSV overview; filing into a DMS is a copy step.
Sorting the mail takes longer than scanning it
Scanning a stack takes a couple of minutes. Deciding what each page is and where it goes takes the rest of the morning.
Open, read, decide, file
Every letter passes through your hands once: open the envelope, skim it, decide where it belongs. With 20 letters, half an hour is gone.
One scan, one filename: scan_0412.pdf
The batch scan has no names inside. To know which letter sits on page 17, someone has to open the file and page through it.
Monday's invoice is missing on Friday
Without a fixed routine, mail sits in a tray. Finding one letter means flipping through stacks or asking around, and deadlines slip.
Docusplit in numbers
The facts before you sign up.
The app's own estimate; a 30-page scan takes a few minutes
Docusplit product dataBy hand or with a batch scan
What changes in the daily routine. The minute figures are the assumptions from the cost example.
| Step | By hand | With Docusplit |
|---|---|---|
| Time for 20 letters | about 30 minutes (assumed 1.5 minutes per letter) | about 5 minutes hands-on; the AI runs around 5 minutes on its own |
| Filename | scan_0412.pdf, or typed by hand | Notice_City-Utilities_2026-03-15.pdf, automatic |
| Handoff to the DMS | entered letter by letter | copy the files, CSV for double-checking |
| Email attachments | separate routine | uploaded like a scan |
| Unreadable pages | noticed while reading | named Document_Page_N.pdf and listed as "Not recognized" in the CSV |
The daily routine in three steps
Scan, upload, unzip. The sorting is the AI's job.
Scan the day's mail as one PDF
Open the envelopes, remove staples, run the whole stack through in one pass. A sheet-fed scanner feeds it double-sided; a multifunction printer or a phone scanning app works too.
Upload and pick Documents mode
The AI reads each page, spots where a new letter begins, keeps multi-page letters together and names each file by type, sender and date. Expect about 10 seconds per page; progress is shown live.
Download the ZIP and file it
You get one PDF per letter, optionally in folders by month or by sender, plus a CSV overview. Unzip, copy into your filing system or your DMS import folder, done.
What changes in the morning
Your hands-on work shrinks to scanning, uploading and unzipping.
A routine instead of a tray
Scan and upload once a day. While the AI works you do something else; afterwards the mail sits named in the ZIP.
Files your DMS can take as they are
One letter per file, ISO date, sender in the name. Copy the files into your filing folder or your DMS import folder and use the CSV to double-check.
Email attachments in the same run
A PDF attachment is the same thing to the AI as a scan. Upload it on its own or append it to the day's scan first.
The hands-on work ends after scanning
Incoming mail rarely fails at the scanning step. A sheet-fed scanner pulls the stack through in one pass, and a multifunction printer isn't far behind. The time goes afterwards, into taking the file apart, renaming and filing.
The upload moves that work to the AI. It reads each page, decides where a new letter begins and assigns the filename. In Documents mode the app estimates around 10 seconds per page, so a 30-page scan is done after about five minutes, with progress shown live.
Your own time comes down to three actions, namely scanning the stack, uploading the PDF and unzipping the result. If you scan at the same time every day, there is no tray left on the desk in the evening.
The AI splits by content and names by type, sender and date
You scan the day's mail as one PDF and upload it. The AI reads each page and recognizes where a new letter begins, even when invoices, contracts, notices and advertising are mixed together. No separator sheets or barcodes are needed. Continuation pages are recognized from content, such as "Page 2 of 3" or a letter that carries on, and stay in one file.
Each file gets a name from its content. Invoices are named InvoiceNumber_Company, for example INV-2026-0481_City-Utilities.pdf. Everything else is named Type_Sender_Date, for example Notice_City-Utilities_2026-03-15.pdf or Contract_Northwind-Ltd_2026-02-28.pdf. The date is written as YYYY-MM-DD so files sort chronologically.
Only the fields for the filename are read, that is document type, sender, date, and for invoices the number and company. Amounts, bank details or line items are not extracted; that remains the job of your accounting software.
ZIP and CSV are the handoff to your DMS; the import stays with you
After processing you download a ZIP. Inside are the separated PDFs, either flat or in folders by month (2026-03/) or by sender. Pages without a recognizable date go into an Unknown/ folder. An _overview.csv and a _metadata.json come with it.
The CSV lists, per file, the filename, page, document type, sender, date, invoice number and company where applicable, and a confidence value as a rough guide. Fields the AI could not read are marked as "Not recognized". It is an overview for double-checking, not an accounting export.
There is no connector to SharePoint, DocuWare, M-Files or any other DMS. You unzip the archive into your DMS import folder or into the sync folder of Google Drive, OneDrive or Nextcloud, and the folder structure is kept. The ZIP does not replace an archive that enforces immutability.
If a virtual mailbox or a scanning service already delivers your mail as PDFs, the same upload applies.
Any scanner will do, even a phone
You don't need special hardware for the upload. The AI finds the boundaries between letters in the content, not in separator sheets or barcodes. An office printer with a scan function, a sheet-fed scanner or a scanning app on your phone is enough, as long as a legible PDF comes out at the end.
A sheet-fed scanner with an automatic document feeder pays off when a lot of mail arrives every day, because it pulls the whole stack through in one pass, both sides. For occasional mail, the multifunction printer is fine.
Scan at around 300 dpi, in color if letterheads and stamps matter. That keeps the file legible without making it huge; the limit is 50 MB per upload. Scans well below 200 dpi, handwriting and very faint pages are harder for the AI to read; you will find such pages as Document_Page_N.pdf.
What a filename does for record-keeping, and what it doesn't
Once business mail exists only as a scan, your record-keeping rules apply to the scan. How long letters, invoices and contracts must be kept differs by country and by document type, so check the rules that apply to you; for US readers, the IRS general rule for tax records is 3 years, with some cases stretching to 7. The upload does not decide that for you.
What the upload gives you is the groundwork. One document per file, a sortable filename with an ISO date, the sender in the name, and a CSV that lists what the AI read from each file. That makes retrieval easier and makes it obvious which files still need a look.
It does not give you an archive that prevents later changes, a documented process or a legally recognized audit trail. Those remain the job of your DMS and of your own procedures. If you receive electronic invoices with an embedded XML part, keep the original file; splitting preserves only the visible PDF pages.
A cost example with the assumptions spelled out
Assumptions, not a study: 20 letters a day, 250 working days, 1.5 minutes per letter by hand for opening, sorting, scanning and naming, and an hourly rate of $30 as a placeholder for your own figure.
By hand, 20 letters take 30 minutes a day, about 125 hours a year, roughly $3,750.
With a batch scan and upload, scanning, uploading and unzipping take about 5 minutes a day, around 21 hours a year (about $630). Add the plan: 20 letters at an average of 1.5 pages make roughly 30 pages a day and about 600 pages a month, which is the Max plan at €99.99 a month for 1,000 pages, around €1,200 a year. Together that is about $630 plus roughly €1,200. The AI's processing time doesn't count, because you do something else while it runs.
Whether this pays off for you depends on the volume and on how long sorting really takes today. The 30 free pages let you test the assumption with your own mail.
What ends up in the ZIP
The features that matter for incoming mail, without embellishment.
- One upload for the whole day's mail, a single PDF up to 50 MB; page count limited only by your quota
- Documents mode for mixed mail with invoices, letters, contracts, notices and receipts
- Filenames from content as Type_Sender_Date, or InvoiceNumber_Company for invoices
- Multi-page letters stay together; continuation pages are recognized from content
- Optional folders in the ZIP by month (2026-03/) or by sender
- CSV overview with filename, page, type, sender, date and confidence, plus _metadata.json
Three inboxes, one routine
Wherever mail arrives daily and should be filed the same day.
Law firm front desk, 8:30 a.m.
Twenty letters, three of them court notices with deadlines. The stack is scanned once; after the upload every letter is named by sender and date, and the notices stand out in the CSV.
Property manager with 40 buildings
Tenant letters, utility invoices, insurance mail. With folders by sender, each sender's letters land in one folder, and invoices are named by invoice number and company.
Distributed team, office mail once a week
Whoever passes by the office scans the week's stack and uploads it. The ZIP goes into the shared folder; everyone finds their letters by filename.
Questions about sorting scanned mail
How long does a 30-page day's mail take?▼
The app estimates about 10 seconds per page in Documents mode, so a 30-page scan takes around five minutes. Progress is shown live and you can do something else meanwhile. Your own time is limited to scanning, uploading and unzipping.
Can I import the CSV into my DMS?▼
The CSV is an overview with filename, page, document type, sender, date and confidence. Whether your DMS can read it depends on the system; there is no connector to SharePoint, DocuWare or M-Files. In practice you copy the named PDFs into your DMS import folder and use the CSV to double-check.
Does it work with email attachments?▼
Yes. A PDF attachment is the same thing to the AI as a scan. Upload it on its own or append it to the day's scan first. Only image files such as JPG or PNG need converting to PDF beforehand.
What happens when the AI splits wrongly or can't read a page?▼
Handwriting, very faint pages, scans below roughly 200 dpi, blank back sides or attachments without a letterhead can be assigned wrongly. If the AI reads nothing on a page, the file is named Document_Page_17.pdf and, with folders by month, lands in Unknown/. Check those files and the confidence column in the CSV first; the rest of the stack is unaffected.
What happens to my files?▼
Files are transferred over TLS, kept only while processing and then deleted from our servers. AI analysis runs through the OpenAI API (data processing agreement with OpenAI Ireland, Standard Contractual Clauses, never used for training; OpenAI retains API data for up to 30 days for abuse monitoring). See our privacy policy.
Related guides
How the AI splits and names, in detail.
Split & Rename PDF
Upload one batch scan of mixed invoices, letters and receipts. The AI splits it into separate documents and renames each file by content. Sorted ZIP with CSV.
Read guideAuto Rename PDF Files
The AI reads type, sender and date on each scanned page and names the file after it, e.g. Contract_Northwind-Ltd_2025-01-15.pdf. 30 pages free, no credit card.
Read guideSources & References
- OpenAI — Your data (API platform) — Data sent to the OpenAI API is not used to train OpenAI models; abuse-monitoring logs are retained for up to 30 days.
- Adobe Community — Scanning multiple documents and renaming them by content — Thread asking Acrobat to read a value from each scanned document and rename the files; the answers agree Acrobat has no native content-based renaming and point to manual bookmarks or custom scripts.
- IRS — How long should I keep records? — The general rule is to keep records for 3 years; 7 years for claims involving worthless securities or bad debt, and employment tax records for at least 4 years.
Today's mail, filed today
Upload the next day's scan and look at the ZIP. The first 30 pages are free.
Try free now30 free pages · no credit card · deleted after processing
Last updated: