Document OCR V2 is a new extraction step in Workflow Builder. It reads your document page by page and carries what it learns on each page into the next. A line-item table that runs across five pages comes out as one table, with every row in order. If you stitch invoice tables back together by hand after extraction, you can stop.
Kessler SupplyINDUSTRIAL · DAYTON OH
Commercial InvoiceNo. INV-20418Nov 3, 2026 · PO-88215Page 1 of 3
16Hydraulic hose assembly, 1/2 in × 36 in, SAE 100R2181–2769.50
17Hydraulic hose, 3/4 × 48 in1821,047.60
18Gear pump, 11 GPM22972.00
19Solenoid valve 24 V DC82594.40
20Pressure gauge 0–3000 psi102386.00
3 pages→1 table52 rows, in order
How it works
Split Document and OCR reads pages in parallel. That’s fast, but it reads each page without knowing anything about the pages around it. When a table continues past a page break, the rows on page two lose their column headers, and you get one short table per page. Document OCR V2 reads pages in order instead, so page three always knows what page two found. You’ll find it next to Document OCR and Split Document and OCR in the step picker.
After each set of pages, Document OCR V2 tracks the current document, the labels it has found, and any table that’s still open: its column layout, its last few rows, and its row count. The next pages either continue that table and document or start a new one. So a continuation page with no header row still maps values to the right columns. A header repeated on every page doesn’t turn into a data row. A row whose description wraps onto the next page stays one row. Labels work the same way: the vendor and PO number from page one and the Total Due on the last page land in the same document. Separate tables on the same page, like line items and additional charges, stay separate.
You can pick more than one schema under Document Types, and Docxster works out which one fits each document. If a PDF has an invoice on page one and a purchase order on page two, each comes out under its own schema. Turn on Support Multiple Instances to split a file of several invoices into one document per invoice, each with its own page range. A 30-page file of back-to-back invoices comes out as 13 invoices, each with its own table and totals. When you pass in more than one file, Document OCR V2 reads each one on its own, so rows from the end of one file never spill into the next.
Reading pages in order takes longer than reading them all at once, so the work runs in the background. Your flow run shows as Processing while extraction is underway and picks back up as soon as it finishes. Long documents don’t hit the usual step time limit.
Each document can be up to 50 pages. A 50-page invoice with 1,000 line items comes through as a single table. The output uses the same structure as the other OCR actions, so later steps like Export to Spreadsheet take it without any changes. You can turn on Support Validation Platform and Enable Human Review the same way you do today.
Where this is useful
Freight and logistics operations
Packing lists and carrier invoices tend to run long, and a shipment file often bundles several of them into one PDF. Turn on Multiple Instances and each document comes out separately with its own page range. Every multi-page table inside it stays whole, so a long freight invoice arrives as one table, not a stack of page fragments.
Customs brokers
Commercial invoices on an entry often run past page one, with HS codes, quantities, and declared values on every line. Document OCR V2 keeps each continuation row tied to the right HS code and value columns. The full item list arrives as one table, ready for classification. Turn on Enable Human Review if you want your team to check the lines before they move on.
Lending and mortgage
Bank statements in a loan file often run many pages, and one missing or duplicated row throws off every balance after it. Document OCR V2 reads page by page, so transactions stay in order with no gaps or repeats. The totals on the last page sit in the same document as the account details from page one, so each statement stays complete for income verification.
Manufacturing and procurement
Supplier invoices for large orders can carry hundreds of lines, and suppliers often send them in the same file as the purchase order. Pick both Invoice and Purchase Order under Document Types, and Docxster sorts each document into the right schema. Line items on both come through complete, ready for PO matching.
On This Page
No headings found
Turn documents into decisions.
See how Docxster gets you from inbox to insight in minutes, not days. Bring your toughest workflow. We'll show you what it looks like solved.