Capture, OCR & indexing
How this is scored
How documents get in and become findable: scanning and email intake, OCR, classification and metadata extraction, and whether incoming e-invoices (XRechnung, ZUGFeRD) are read as structured data rather than as pictures.
0 — Manual upload only; no OCR, no metadata beyond file name and folder.
3 — Upload, drag-and-drop and scan import with OCR mentioned, but indexing is manual and no word on email intake or e-invoice formats.
5 — Scanner, email and folder intake, full-text OCR, document types with metadata fields, and automatic extraction of common fields (date, sender, amount) documented.
8 — AI- or rule-based classification and field extraction with confidence and a validation step, XRechnung and ZUGFeRD read as structured data, mailbox and Microsoft 365 intake, and duplicate detection documented.
10 — Capture is engineered and shown: classification and extraction trainable per document type, validation queues, e-invoice validation against the standard, intake channels listed with their limits, and where AI processing runs (location, model provider) stated.
The Skeptic
Intelligent document processing is documented in substance: classification, extraction with confidence scores, a trainable custom model fed a few sample documents, verification of extracted bank details, OCR/HTR and zero-shot GenAI extraction, plus e-invoices that are received, validated and made PEPPOL-fit. But we found no public information on intake channels and their limits, on mailbox or Microsoft 365 ingestion, or on duplicate detection, and despite all the GenAI talk no word on where that AI processing runs. Strong on the reading side, thinner on the intake side. 1 4 5 7 10 11