# How to add document OCR to IFS Cloud > Source: https://netray.co/integrations/ifs-cloud-document-ocr > Publisher: Netray (netray.co), AI and integration engineering for ERP and manufacturing systems. > Section: Integrations ## Short answer To add OCR to IFS Cloud, route incoming documents to a capture service that extracts fields with confidence scores, matches them to IFS purchase orders and suppliers, and stages the result for a person to approve. Approved data is posted through projections or IFS Connect, with the image stored as a document on the record. ## System IFS Cloud ## Capability document-ocr ## Intro Accounts payable, receiving and quality teams on IFS Cloud still process paper-shaped documents: supplier invoices, packing slips, certificates of conformance, material test reports, delivery notes and customer drawings. The ERP is strong on the transaction, but the step before it, reading the document, is manual in many plants. Capture products exist in the IFS ecosystem, and IFS partners offer invoice scanning. Whether those fit depends on volume, document variety and where the documents may be processed. Manufacturers in regulated sectors often cannot send drawings and certificates to a public cloud service, which pushes them toward local OCR and open-weight models. This page lays out the architecture, what to extract per document type, how to use confidence thresholds, and where document capture projects fail in practice. ## Use Cases ### Supplier invoice capture to matched supplier invoice **Description:** Read header and lines, match to the purchase order and receipt, and stage a supplier invoice for the AP clerk with differences highlighted. ### Packing slip to receipt **Description:** Extract PO number, part numbers and quantities from a supplier packing slip so the receiver confirms instead of keying. ### Certificate of conformance filing **Description:** Read heat, lot and specification references from certificates and attach them to the receipt and lot records. ### Material test report checks **Description:** Extract chemical and mechanical values and compare them against the specification before the lot is released from inspection. ### Customer PO and drawing intake **Description:** Read the customer PO and register the drawing revision, flagging when the revision differs from the part in IFS. ### Freight and customs documents **Description:** Capture commercial invoices and airway bills for shipment records, with HS codes checked against master data. ## System Surface - Projections: create or update supplier invoice, receipt and quality records via the supported OData REST layer, using entity names confirmed in API Explorer for your release. - IFS Connect: a robust channel for posting captured data with queues, retries and routing, useful for batch imports. - Document Management: store the source image or PDF as a document connected to the order, receipt, invoice or lot through document reference objects. - Master and transaction lookups: suppliers, purchase orders, purchase receipts, part catalog and inspection plans are the data the matcher relies on. - Partner and IFS-delivered capture options may exist for supplier invoices in your edition, so review them before building. ## How It Works ### Define document types and fields **Detail:** For each type, list fields, validation rules and which IFS record it must link to. Start with one type, usually supplier invoices or packing slips. ### Ingest documents **Detail:** Take files from a scan folder, shared mailbox, supplier portal or scanner app into a queue, with a hash to prevent duplicates. ### Run OCR and layout analysis **Detail:** Use a local OCR engine plus a layout-aware model for tables. Keep the page images and coordinates for the review screen. ### Extract and score **Detail:** Produce fields with a confidence value each, and run business validations such as totals adding up and tax rules. ### Match to IFS **Detail:** Find the supplier, purchase order and receipt lines, and compare quantities and prices to tolerances. ### Review screen **Detail:** Show the image beside the fields, highlight low confidence and mismatches, and let a clerk correct and approve. ### Post and archive **Detail:** Create the record through the projection or Connect message, attach the document and store the extraction trail for audit. ## Example Flows ### Three-way match for a supplier invoice **Trigger:** A PDF invoice arrives in the AP mailbox **Steps:** - File enters the queue and duplicate hash check passes - OCR and extraction produce header and lines with confidence scores - Matcher finds the PO and receipts and computes variances - Clerk reviews highlighted price and quantity differences - Approved invoice is created through the projection with the PDF attached **Result:** Clerks resolve exceptions rather than type every line, with the document linked for audit. ### Certificate attached to a receipt **Trigger:** A supplier sends a certificate of conformance with a delivery **Steps:** - Receiving scans the certificate - Extraction reads lot, heat number and specification - Matcher links it to the receipt line and lot - Missing or mismatched specifications raise a quality task **Result:** Traceability documents are filed on the lot at receipt time. ### Packing slip to receiving **Trigger:** A delivery arrives at the dock **Steps:** - Receiver photographs the packing slip on a mobile device - Capture reads PO and quantities - Receipt lines are pre-filled for confirmation - Differences are marked for the buyer **Result:** Receiving transactions are faster and shortages are noticed at the dock. ## On Prem Capture is where on-prem matters most, because invoices show pricing and certificates and drawings can be controlled data. Run OCR and the extraction model on your own GPU servers or in a private cloud, using open-weight vision-language and text models served with vLLM or Ollama. The capture service talks to IFS Cloud only through projections or IFS Connect on your network. Keep scanned images in storage you control, and use IFS document management for the record copy. Accuracy should be proven on your own documents before go-live, especially on poor scans and multi-column tables. ## Security - Encrypt documents at rest and in transit, and restrict the review screen by role. - Segregate duties: the person approving extraction should not be the only approver of payment. - Verify bank detail changes out of band; never auto-accept changed remittance details from a document. - Keep a full extraction log with model version and corrections for audit. - Apply retention rules to scan images that match your records policy. ## Options ### IFS-delivered or partner invoice capture product **Fit:** Standard supplier invoice processing at scale **Effort:** Weeks to a few months with vendor configuration ### Build capture with local OCR and open-weight models **Fit:** Varied documents, controlled data, many types beyond invoices **Effort:** Roughly two to four months for the first type ### Partner delivery with Netray plus your IFS partner **Fit:** Capture plus matching, review UI and IFS posting, on-prem **Effort:** Phased, one document type per phase ## Pitfalls - Judging accuracy on clean sample PDFs instead of your worst scans and faxed copies. - Posting before the three-way match, which pushes errors into payables. - Failing to detect duplicate invoices re-sent by email and by post. - Not tracking correction rates by supplier, which hides the few formats causing most rework. - Treating multi-page tables and credit notes as an edge case when they are common. ## Faq ### Does IFS Cloud include invoice OCR? **Answer:** IFS and its partners have offered supplier invoice capture options, but what is included depends on your edition and contract. Check with your IFS account team. For other documents like certificates and packing slips, a custom or partner capture layer is more common. ### What confidence threshold should we use? **Answer:** Set it per field, not per document. Amounts and PO numbers need a strict threshold with mandatory review below it. Descriptions can be looser. Begin with full review, then relax thresholds only for suppliers whose measured correction rate is low. ### Can OCR run fully on-prem? **Answer:** Yes. Open-source OCR engines and open-weight vision-language models run on your own GPUs. Quality varies with scan quality and layout, so evaluate on a few hundred of your own documents and include handwritten notes and stamps in the test. ### How do we link the scanned file to IFS? **Answer:** Attach it through IFS document management as a document connected to the invoice, receipt or lot, using the supported projections or document services for your release. Keep the capture system's id on the record so you can trace back. ### Which document type should we start with? **Answer:** Start with the one with the highest volume and a clear matching key, usually supplier invoices or packing slips. Both match to purchase orders, so you can measure time saved and error rates, then add certificates and test reports. ## Search terms IFS Cloud OCR, IFS Cloud invoice capture, IFS supplier invoice OCR, IFS document scanning, IFS Cloud certificate of conformance OCR, IFS Applications invoice automation, IFS packing slip OCR, AI document extraction IFS, IFS Cloud AP automation --- Talk to an engineer: https://netray.co/contact?intent=erp_ai | Machine catalog: https://netray.co/agent/catalog.json