Rossum Paid

Document AI for invoices and purchase orders that learns each transaction layout

What is Rossum

Rossum is document AI for transaction paperwork: it reads invoices, purchase orders, credit notes and packing lists, extracts the fields, and validates them against your ERP before they are booked.

The Prague company started in 2017 with two founders who previously built OCR tools for the security industry. Its early pitch was unusual at the time: instead of a per-page template per supplier, Rossum learned how each supplier formats a document so that the next invoice from the same vendor needed no setup.

Since 2024 it has shifted the extraction engine to LLMs, keeping the original validation layer. That combination matters more than raw OCR accuracy, because a wrong line item on an invoice is a payment error, not a typo.

Product Features

  • LLM-based extraction — handles scanned PDFs, photos and native files across roughly 35 languages without per-vendor templates.
  • Validation rules — checks totals, VAT rates, purchase order references and line sums, flagging mismatches before export.
  • Human-in-the-loop queue — low-confidence fields are routed to a review screen; reviewers accept or correct, and corrections retrain the model for that document type.
  • ERP connectors — SAP, Oracle NetSuite, Microsoft Dynamics 365 Business Central, Coupa and Basware have prebuilt integrations.
  • API-first — every extraction is exposed as a REST endpoint, which is how most customers embed it in their own invoice pipeline.
  • Pricing is quote-based — no public per-document list price; expect annual contracts sized on document volume, sold as Enterprise and Growth tiers.
  • Accuracy claims need testing — Rossum advertises high field-level accuracy, but the number depends entirely on your supplier mix, so request a pilot on your own PDFs.
  • Audit trail — keeps the original file, the extracted JSON and the reviewer decisions for compliance export.
  • Queue workflows — routes documents across teams by country, entity or value, useful for shared service centres.
  • No free plan — only a scoped pilot or trial arranged with sales.

Application Scenarios

  • Accounts payable automation — booking invoices without a human typing line items, with the exception queue catching bad scans.
  • Shared service centres — one team processing invoices in several countries and languages with entity-level routing.
  • Purchase-to-pay — matching purchase orders and goods receipts against invoices before a payment run.
  • Collections and logistics — freight documents and packing lists use the same extraction pipeline as invoices.
  • Where it does not fit — small businesses processing under a few hundred invoices a month. Quote-based pricing and an implementation project rarely pay back at that volume, and tools like Nanonets or Airparser are easier to self-serve.

Quick Summary

  • Rossum is Prague-based document AI, founded in 2017, focused on invoices and transaction paperwork.
  • Extraction is LLM-driven with a human review queue and ERP integrations for SAP, NetSuite and Business Central.
  • Pricing is quote-based with annual contracts; there is no free tier or published per-page rate.
  • Built for AP teams handling thousands of documents a month, not for low-volume self-serve users.