How to add document OCR to Infor LN
Short answer
Add OCR to Infor LN by sending scanned and PDF documents to an extraction service, validating the output against LN masters and open orders, and posting accepted results through ION or an LN business interface. Low-confidence fields go to a review screen, and the source image is attached to the LN record.
LN holds the structured transaction, but the evidence for that transaction is paper and PDF: supplier invoices, packing slips, certificates of conformance, test reports, inspection records and customer drawings. In many plants these still get keyed by hand or scanned into a folder with no link to the order.
Modern extraction can read these documents with good accuracy, but accuracy alone is not enough. What matters is matching to the right LN receipt, purchase order or lot, handling partial shipments, and knowing when to ask a person. A flow that is 95 percent right and quiet about the other 5 percent is worse than a slower one that flags uncertainty.
This page covers the LN touchpoints for invoice and receiving capture, a design for confidence and review, and options from a packaged AP product to a private on-prem pipeline.
What teams use it for
Supplier invoice capture and matching
Extract invoice header and lines, then perform a match against purchase order and receipt in LN before creating the purchase invoice entry for approval.
Packing slip to receipt
Read the packing slip at the dock, match to the open PO, and prefill the receipt with quantities and lot numbers for the receiver to confirm.
Certificates of conformance and mill certs
Extract heat or lot numbers and specifications from certificates, link them to the received lot in LN, and flag mismatches with the purchase specification.
Customer drawing and spec intake
Classify incoming drawings by revision and part number, then attach them to the relevant item, quotation or sales order with revision checks.
Quality inspection records
Digitise handwritten or scanned inspection sheets into measurement values for quality orders, with a review step on illegible entries.
Infor LN integration surface
- Infor ION API Gateway and BODs for posting purchase invoices, receipts and related documents, subject to version and API set availability.
- LN Accounts Payable and invoicing sessions, including three-way matching against purchase orders and receipts, as the validation backbone for invoice capture.
- LN warehousing receipt processes, where receipt lines are tied to purchase order lines and inbound advice, used for packing slip matching.
- Infor Document Management or LN attachments to store the source image with the transaction.
- File-based import via Exchange schemes for installations without ION, and read replica access for matching lookups.
How it works
- 1
Collect a representative sample
Gather a few hundred real documents from your top suppliers and note scan quality, languages, tax formats and multi-page layouts.
- 2
Choose capture and OCR engine
Decide between a packaged invoice product, a cloud extraction API, or a locally hosted OCR and vision language model. Test on your sample before committing.
- 3
Extract with confidence scores
Return every field with a confidence, plus page and region references so a reviewer can see where the value came from.
- 4
Validate against LN
Match supplier, PO, receipt, item and tax codes. Check totals and line sums. Block anything that fails arithmetic or master data checks.
- 5
Review exceptions
Present only failed or low-confidence fields beside the image. Corrections become training data and rule updates.
- 6
Post and attach
Send accepted documents to LN via ION or a business interface into a pending or approval status, with the source image attached and the extraction log stored.
Example flows
Invoice with three-way match
Trigger: A supplier invoice PDF arrives in the AP inbox or scanner folder.
- The service extracts supplier, invoice number, dates, tax and lines.
- It finds the PO and the matching receipts in LN and compares quantities and unit prices within tolerance.
- Matched invoices are posted as pending approval; variances go to the AP clerk with the difference highlighted.
Result: Clerks handle exceptions, not every invoice, and each posted invoice carries its source image.
Certificate of conformance at receiving
Trigger: A certificate is scanned together with the packing slip.
- OCR reads lot or heat number, material grade and standard.
- The service compares them with the PO specification and the received lot.
- On a match, the certificate is attached to the lot in LN; on a mismatch, the lot is held for quality review.
Result: Traceability records are linked at receipt, and mismatches are caught before material reaches production.
Keeping data on-prem
Document capture is a strong on-prem candidate because the documents contain pricing, customer part numbers and sometimes controlled drawings. Run OCR and a vision or language model on your own GPUs with open-weight models, keep the validation service next to LN, and post through ION inside the network. Cloud extraction APIs are simpler to start with but send full documents outside your environment, which may conflict with contractual or export obligations. A hybrid option is to use local processing for sensitive supplier classes and a hosted service for routine invoices.
Security checklist
- Encrypt documents at rest and in transit, and limit retention of working copies once the LN record holds the attachment.
- Do not send controlled drawings or technical data to a hosted OCR service unless your compliance team approves.
- Use a dedicated integration user and require approval before invoices are released for payment.
- Detect duplicate invoices and changed bank details as fraud signals and route to review.
Ways to get it done
Packaged AP automation product
Effort: Weeks to months including connector setup, with licence fees per document.
Good if you mainly need invoices and the vendor supports LN or ION posting.
Build with an OCR engine and your own validation
Effort: Several weeks per document type.
Suits teams wanting control over matching logic and multiple document types.
Partner build such as Netray with private models
Effort: A pilot on one document type, then extension.
When documents must stay on-prem or you need certificates and drawings handled alongside invoices.
Pitfalls
- !Measuring accuracy by character instead of by whole documents that post correctly.
- !Ignoring partial deliveries and multiple invoices per PO during matching.
- !Letting the model guess item numbers; require an exact match or review.
- !Skipping the image attachment, which weakens audit trails and dispute handling.
Frequently asked questions
Does Infor LN include OCR?
Infor offers document and invoice capture capabilities in its wider portfolio, and availability varies by contract and version. Many LN customers use third-party capture tools or custom services posting via ION. Check what your licence already covers before buying.
How do we post captured invoices into LN?
Typically through ION APIs or BODs, or LN business interfaces, into a status that needs approval. Older installations may use file import schemes. Direct database writes should be avoided because they bypass LN validation.
Can OCR handle handwritten documents?
Printed text is far more reliable than handwriting. Modern vision models can read neat handwriting, but critical values such as quantities or measurements should always get a human check with the image visible.
What confidence threshold should we use?
There is no universal number. Start strict, route most documents to review, and relax thresholds per field and supplier once your correction data shows consistent accuracy. Totals and identifiers deserve tighter limits than descriptions.
Can this run entirely on our own servers?
Yes. OCR and open-weight vision or language models can run on local GPUs, with validation and posting inside your network. Throughput depends on hardware, so size it against your daily document volume.
Want Document OCR working on your Infor LN?
A free 30 minute assessment: we map the integration path for your version and setup, and where it should run.
Get a free assessment