Any inbound PDF — vendor invoices, sales orders, RFQs, freight bills, etc. — extracted and posted to your ERP.

Vendor invoices, customer sales orders, email-attached invoices, RFQs (converted to quotes), freight/carrier invoices, credit memos, and order acknowledgements. An AI classifier identifies the document type and routes each one to the correct ERP path — documents that are ambiguous or unrecognized land in a Needs Review queue rather than passing through unchecked.

P21 (Prophet 21), several Infor systems (including Infor SX.e), NetSuite, and others. Each document type routes to the correct API call for that ERP — for example, a vendor invoice on P21 posts via ReceiveP21VendorInvoice, while a customer order creates a sales order via CreateP21SalesOrder.

No. AI Document Intake uses Claude to read the PDF and return structured data — vendor name, invoice number, PO reference, line items, totals, freight, and tax — without any per-vendor template. If the AI extraction looks weak (spam classifier triggers, missing invoice number or line items), the pipeline automatically retries with a more capable model as a failsafe.

The pipeline runs a duplicate invoice check, an arithmetic check (lines + freight + tax = total), PO line matching against the ERP, and item ID lookup. Documents with no vendor match are flagged and an email is sent to your operations team — the pipeline does not silently create phantom vendors.

Via email attachments, a watched drop folder (PDFInvoiceDirectory), FTP, AS2, or HTTP upload — the same intake channels your EDI traffic already uses. There is no separate portal or manual upload step required.

It fills the gap EDI doesn't cover. EDI handles trading partners who exchange structured electronic transactions — 810 invoices, 850 purchase orders, 856 ship notices. But some vendors, particularly smaller suppliers, carriers, and international partners, send PDFs instead. PDF Import processes those PDFs through the same intake channels your EDI traffic already uses — AS2, FTP, email, drop folder — and posts them to your ERP the same way. Your operations team sees a consistent flow of posted transactions regardless of whether the source was an EDI transaction or a vendor PDF.

OCR reads text from a document but doesn't understand what that text means. Template-based systems require you to define where each field sits on each vendor's layout — and break when a vendor changes their format. PDF Import uses AI to understand the document semantically, extracting the right fields regardless of layout with no templates to build or maintain. Real examples where this matters: ship-to addresses spanning four or five lines; order lines that break across page boundaries; quantities appearing in non-standard positions; European number formats where a period is the thousands separator and a comma is the decimal point (e.g. 1.234,56); and date formats that vary by country or region. Each of these would require manual correction with OCR — the AI resolves them automatically.

Yes. Image-based PDFs — scanned documents where the pages are rasterized images rather than selectable text — are fully supported. The AI detection step identifies the document type and extracts data regardless of whether the PDF is text-based or image-based. No separate OCR pre-processing step is required; image handling is part of the same AI classification that checks whether a document is a vendor invoice, sales order, carrier invoice, or something else entirely.