How to add document OCR to Infor M3
Short answer
To add OCR to Infor M3, capture documents into a processing service that extracts fields, scores confidence, matches them to M3 data, and posts approved results through ION or M3 APIs, with the source file stored in Infor Document Management. Invoices are the usual first target, followed by packing slips and certificates. Low-confidence fields always go to human review.
Document capture in M3 environments is often a mix of generations: a legacy scanning product for invoices, manual keying for packing slips, paper certificates filed in a binder. Finance teams re-key supplier invoices, warehouse staff type receipts against purchase orders, and quality teams chase certificates of conformance for traceability.
OCR alone is a commodity. What matters in M3 is what happens after reading: matching an invoice to purchase orders and receipts, resolving suppliers and items, coding to the right accounts and dimensions, and creating a record with the right approval. That means understanding M3's invoice matching, receiving and supplier data, not only text extraction.
The goal is a pipeline where clean documents flow through, uncertain ones get a quick look, and every posted transaction links back to the image that justified it.
What teams use it for
Supplier invoice capture and matching
Extract header and line data from PDFs, match to purchase orders and goods receipts, and create the supplier invoice batch for review in the invoice matching workflow.
Packing slip to goods receipt
Read supplier packing slips, resolve the PO and lines, and prefill the receipt quantities for the warehouse to confirm.
Certificates of conformance and mill certs
Extract heat or lot numbers, material specifications and test values from certificates and attach them to the lot or item receipt for traceability.
Customer PO attachments
Read scanned customer orders from mail or portals and prepare a draft customer order for sales administration.
Drawings and revision control checks
Read title block data (drawing number, revision) from supplier or customer drawings and compare with the revision on the M3 item or product structure.
Delivery documents and proof of delivery
Capture signed delivery notes, link them to the delivery in M3 and mark proof of delivery received.
Infor M3 integration surface
- Infor Document Management (IDM) and Infor OS document services: store source images and link them to entities such as supplier invoices, POs or items. Entity names and attribute setup vary by installation.
- M3 API (MI programs) for creating and reading purchase orders, receipts, suppliers and invoice batches. Invoice matching and AP programs are version-dependent, so verify names in the API metadata.
- ION API gateway and ION Connect/BODs: use ION to receive document-related events or to push documents into IDM with metadata.
- Existing invoice capture or scanning tools often hand off through a file drop or an M3 interface (a CSV/XML interface via MEC or an M3 import). Replacing only the extraction step lets you reuse that hand-off.
- M3 master data (suppliers, items, supplier-item cross references, GL accounts and dimensions) used to validate and code extracted fields.
How it works
- 1
Choose the document types by value
Rank by volume, effort and risk. Invoices usually win on volume; certificates win on compliance value. Do not start with every document at once.
- 2
Set up capture channels
Scanner folders, a shared mailbox, a supplier portal upload or a mobile photo. Normalise to PDF or image with a stable document ID and checksum.
- 3
Extract with a layout-aware model
Use OCR plus layout analysis or a vision model to produce structured JSON (header, lines, totals, dates, tax) with a confidence value per field.
- 4
Validate against M3
Check supplier, PO number, currency, line quantities and prices against M3. Flag differences beyond tolerances. Apply arithmetic checks on subtotals and tax.
- 5
Review exceptions only
Display the image beside the fields. Fields under the confidence threshold or failing validation are highlighted. Reviewers correct and approve; corrections feed back into rules.
- 6
Post and archive
Create the M3 transaction through the API or the interface your AP process uses, store the image in IDM with metadata, and record the audit trail.
Example flows
Three-way match for a supplier invoice
Trigger: A PDF invoice arrives in the AP mailbox.
- The service extracts supplier, invoice number, date, PO reference, lines and tax.
- It matches the PO and goods receipts in M3 and compares quantities and prices.
- Clean matches are sent to the invoice matching process; exceptions go to the AP clerk with highlighted fields.
- The image is stored in IDM and linked to the invoice record.
Result: Clerks handle exceptions rather than keying every line, and duplicate invoices are caught early.
Packing slip receiving
Trigger: A warehouse user photographs a packing slip on arrival.
- The service reads the PO number and line quantities.
- It proposes a receipt matching open PO lines, noting over- or under-deliveries.
- The user confirms on a screen, and the receipt is posted via the API.
Result: Faster receiving with fewer mismatches between slip and receipt.
Material certificate capture
Trigger: A mill certificate PDF is uploaded against a received lot.
- The service extracts heat number, specification and key test values.
- It compares the specification with the item's requirement.
- A quality inspector confirms, and the certificate is attached to the lot record.
Result: Traceability from lot to certificate without hunting through paper files.
Keeping data on-prem
Documents often contain pricing, part numbers and controlled technical data, so run extraction where the data already lives. Open vision-language and OCR models can run on a local GPU server, with images stored in IDM or a local file store and posted into M3 through its internal API. This removes the cloud processing step entirely. Expect to tune on your own suppliers' layouts: accuracy varies widely with scan quality, handwriting and the number of distinct formats, so measure on a sample before committing.
Security checklist
- Keep document images inside your network when they contain controlled or confidential data; process with local models.
- Use field-level confidence thresholds and mandatory review for bank details, payment terms and amounts above a limit.
- Detect duplicate invoices by supplier, number and amount before posting, and block changed bank details pending verification.
- Use a dedicated service account for posting with limited MI program access, and segregate duties between capture, review and approval.
- Retain original images and extraction results for the audit retention period required by your finance policy.
Ways to get it done
Dedicated AP automation or capture product with an M3 integration
Effort: Two to four months including AP process changes.
Finance-led invoice automation with vendor support.
Build with open OCR or vision models and M3 APIs
Effort: Two to four months for a first document type.
Mixed document types, need for control over data and costs.
Partner build such as Netray
Effort: A few weeks for a scoped pilot on one document type.
On-prem or air-gapped sites, certificates and non-invoice documents, review UI and measurement.
Pitfalls
- !Measuring accuracy at the character level instead of the field level; what matters is how many documents post without a fix.
- !Skipping validation against M3 masters, which lets plausible but wrong values through.
- !Posting without storing the image link, leaving auditors without evidence.
- !Training only on clean samples; real scans are skewed, stamped and handwritten.
- !Ignoring tax and currency handling on foreign invoices.
Frequently asked questions
Does Infor M3 include OCR?
Infor offers document management and, depending on your licenses and cloud offering, capture or invoice automation features. Check what your contract includes. Many customers use a separate capture product or custom pipeline that posts into M3 using APIs or interfaces.
How do we post extracted invoices into M3?
Either through M3 API programs for supplier invoice or invoice matching entry, or through the existing interface your AP process uses, for instance a file or MEC-based import. The right choice depends on your version and current process.
What accuracy should we expect?
It depends on document quality and layout variety. Machine-printed invoices from repeat suppliers can reach high straight-through rates after tuning, while handwritten or poor scans need review. Measure on a sample of your documents before promising targets.
Can OCR run without sending documents to the cloud?
Yes. Open-source OCR and open-weight vision models can run on local GPUs. You trade some convenience and possibly accuracy on odd layouts for full control of the data, which is often decisive for defense and regulated customers.
Where do images live after posting?
Store them in IDM or another document repository with metadata linking to the M3 transaction. The aim is that anyone looking at an invoice or receipt in M3 can open the original in one click.
Want Document OCR working on your Infor M3?
A free 30 minute assessment: we map the integration path for your version and setup, and where it should run.
Get a free assessment