How to add document OCR to SAP ECC 6.0
Short answer
Extract fields from scanned invoices, packing slips and certificates with an OCR plus language-model pipeline, match them to POs and vendors, and post parked invoices or goods receipt proposals through BAPIs. Store the image with ArchiveLink so the document is visible from the SAP transaction. Low-confidence fields go to a review screen.
Many ECC users still handle paper or PDF invoices manually: scan, key into MIRO or FB60, attach the image. Receiving departments key packing slips into MIGO, and quality departments file certificates of conformance by hand against the batch.
OCR on its own reads text; the real work is understanding the document and tying it to ECC objects. A three-way match in ECC (PO, goods receipt, invoice) is a useful check, because the system already knows what should arrive and at what price.
What teams use it for
Vendor invoice capture
Extract vendor, invoice number, date, totals and lines, match to the PO and park the invoice for approval.
Packing slip to goods receipt
Read delivery note lines, match them to open PO items and propose a goods receipt movement for the dock clerk to confirm.
Certificate of conformance filing
Pull the heat or lot number and part number from certs and attach them to the batch or inspection lot.
Drawing and revision capture
Read title blocks to link drawings to materials and revision levels in the document management system.
Freight and customs documents
Extract values from carrier invoices and customs forms for freight settlement.
SAP ECC 6.0 integration surface
- ArchiveLink (OAC0, OAC2, OAC3) with a content repository stores the image and links it to a business object such as the invoice or PO.
- BAPI_INCOMINGINVOICE_PARK and BAPI_INCOMINGINVOICE_CREATE post vendor invoices; parked documents appear in MIR7 for review and posting in MIRO.
- Goods receipt via BAPI_GOODSMVT_CREATE, with the right movement type (101 for PO receipts, typically) and PO reference.
- SAP DMS (CV01N and related) or Generic Object Services (GOS) can hold certificates and drawings against materials and batches.
- Established invoice-management products in the SAP ecosystem exist; check what you already license before building.
How it works
- 1
Ingest documents
From a mailbox, scanner folder or portal. Keep the original and a checksum.
- 2
OCR and layout
Run OCR with layout detection; for multi-line tables use a layout-aware model, not plain text.
- 3
Extract fields with a model
A language model maps text to a schema (vendor, PO, lines) and gives per-field confidence.
- 4
Match to ECC
Resolve vendor tax IDs and names to LFA1, PO numbers to EKKO, and lines to EKPO with quantity and price tolerances.
- 5
Threshold and review
Above-threshold documents are proposed automatically; others go to a review UI that shows the image next to the fields.
- 6
Post and archive
Park or post through the BAPI and link the image with ArchiveLink to the resulting document.
Example flows
Invoice with a PO
Trigger: A vendor PDF invoice arrives in the AP mailbox.
- OCR and extraction produce header and lines.
- The PO is found and three-way match checks run.
- A tolerance breach is highlighted for the clerk.
- The clerk fixes one quantity and approves.
- The invoice is parked via BAPI and the image archived.
Result: Fewer keying errors and faster approvals, with the image attached to the document.
Certificate for a received batch
Trigger: A supplier cert is scanned at receiving.
- The batch and part numbers are read.
- They are matched to the open inspection lot.
- The PDF is linked to the batch or lot.
Result: Quality can find the cert from the batch.
Keeping data on-prem
Invoices, certificates and drawings are sensitive. Run OCR and the extraction model on your own GPU servers using open-weight vision-language and text models, store images in your content server, and call ECC over internal RFC. This also avoids sending controlled technical data to a public API, a concern for aerospace and defense suppliers.
Security checklist
- Restrict the posting service user to invoice parking and goods receipt for specific company codes and plants.
- Validate bank details changes separately; never let OCR alter vendor bank master data.
- Check for duplicate invoices by vendor, invoice number and amount before posting.
- Keep the original document immutable and log every field edit.
Ways to get it done
Existing SAP-ecosystem invoice product
Effort: Months, vendor-led.
AP-heavy sites with standard invoices and a licence already.
Build with open-source OCR and a local model
Effort: 8 to 14 weeks to a dependable level.
A team that wants control and handles varied documents.
Partner such as Netray
Effort: Pilot on one document type in weeks.
Mixed document types, on-prem requirement and BAPI posting with review.
Pitfalls
- !Tables with merged cells and multi-page invoices break naive extraction.
- !Tax, freight and discount lines do not map to PO items and need rules.
- !Auto-posting without a duplicate check creates double payments.
- !ArchiveLink repository misconfiguration is a common cause of images not appearing in the transaction.
Frequently asked questions
Does ECC have built-in OCR?
Not as a ready AI capability. ECC supports ArchiveLink for storing images and has scanning scenarios through partner products, but extraction and matching are done by add-on tools or custom pipelines.
Should we post or park invoices?
Parking is the safer start. It creates a document that accounting reviews in MIR7 before posting, so errors are corrected inside normal SAP controls.
How do we handle low-confidence fields?
Set a threshold per field. Anything below it is highlighted in a review screen beside the image. Track corrections to tune thresholds and improve prompts.
Can this run fully offline?
Yes. OCR, extraction models, the review UI and the ECC connection can all run on-prem, with no external API calls.
What about handwritten notes?
Handwriting accuracy is much lower and variable. Treat it as review-only and do not auto-post from handwritten content.
Want Document OCR working on your SAP ECC 6.0?
A free 30 minute assessment: we map the integration path for your version and setup, and where it should run.
Get a free assessment