How to add document OCR to ECI M1
Short answer
Scan or receive documents into a watched folder, run OCR and field extraction, match results to M1 purchase orders, receipts and items, and send low-confidence fields to a reviewer. Approved data posts through a supported M1 path with the image attached to the record.
Paper still moves through ECI M1 shops: supplier invoices, packing slips on the receiving dock, certs of conformance and material test reports that must follow a lot, and customer drawings. Each is keyed or filed by hand, and the evidence of what was received lives in a drawer or a network folder with an inconsistent name.
OCR changes the economics only when it is paired with validation. Reading characters is a solved problem; knowing whether the invoice total matches the PO, whether the cert lists the right heat number, and whether a drawing revision is current is what removes work. The design needs thresholds, matching and a fast review screen.
What teams use it for
Receiving from packing slips
Scan the slip at the dock, read PO number, part and quantity, and pre-fill the receipt for the receiver to confirm.
Three-way match for invoices
Extract invoice lines and totals, compare to PO and receipt, and route variances to AP or the buyer.
Certs of conformance and mill certs
Read heat or lot numbers, specification and material grade, and attach the cert to the matching lot with a check against requirements.
Drawing intake and revision check
Read the title block for part number and revision, compare to the item master, and flag mismatches before a job is released.
Legacy paper archive indexing
Bulk-scan old job packets and index them by job or part so they are searchable from the record.
Inspection report capture
Digitize handwritten first article or inspection sheets into structured values for the quality record.
ECI M1 integration surface
- M1 purchase orders, receipts, vendors and items in SQL Server, read-only, as the validation reference. Names vary by version.
- M1's document attachment or linked-document capability, used to store the original image against the record. Check how your release stores them.
- The supported receiving and payables entry interfaces in your release for posting approved data.
- Lot and serial tracking structures, which determine where a cert should attach.
- Network shares or scanner output folders, which are the practical intake point for many sites.
How it works
- 1
Standardize intake
Pick one drop point per document type: a scanner profile, a dock tablet camera or a monitored mailbox, and name files consistently.
- 2
Preprocess images
Deskew, denoise and split multi-document scans. Poor scans cause more errors than model choice.
- 3
Run OCR and extraction
Use a layout-aware OCR engine and field extraction per document type, producing values with per-field confidence.
- 4
Validate against M1
Check PO exists and is open, vendor matches, quantities are within tolerance, and totals reconcile with lines.
- 5
Route by confidence
High-confidence, fully matching documents go to quick confirm; the rest land in a review screen that shows the image beside the fields.
- 6
Post and archive
Post through the supported path, attach the image and log who approved which fields.
Example flows
Dock receiving
Trigger: A truck delivers parts with a packing slip.
- Receiver photographs the slip on a tablet.
- OCR reads PO number and lines and matches the open PO.
- A quantity shortage is highlighted against the PO line.
- Receiver confirms counts and the receipt is posted with the image attached.
Result: Receiving is faster and the slip is permanently linked to the receipt.
Mill cert to lot
Trigger: A material certificate PDF arrives with a bar stock shipment.
- OCR extracts heat number, grade and standard.
- Heat number is matched to the received lot.
- Grade is compared to the item's required specification.
- Mismatches go to quality; clean ones attach to the lot automatically.
Result: Traceability documents are attached at receipt, ready for customer or audit requests.
Keeping data on-prem
Run the OCR engine, extraction model and review screen on servers inside your network, reading from the same file share your scanners already write to. Open-source OCR plus an open-weight vision or language model served locally handles most document types without sending supplier invoices, certs or controlled drawings off-site. A modest GPU server is usually enough for dock-scale volumes, and the review application can be a simple internal web page.
Security checklist
- Keep documents and extracted data on controlled storage with role-based access.
- Do not send export-controlled or customer-restricted drawings to external OCR services.
- Log every field a reviewer changed, with user and time.
- Retain the original image unchanged alongside any extracted values.
- Use a read-only database login for validation and a narrow credential for posting.
Ways to get it done
Commercial IDP or AP automation product
Effort: Weeks, with per-document pricing
Strong for invoices; check that it can post into your M1 version or export a file M1 can ingest.
Open-source OCR and custom matching in-house
Effort: Six to twelve weeks
Workable for a few document types if you can maintain the validation rules.
Partner build with Netray
Effort: Scoped per document type, typically a few weeks each
Useful for certs, drawings and mixed layouts, especially where documents cannot leave your network.
Pitfalls
- !Judging accuracy on clean sample PDFs instead of real dock photos and faxed certs.
- !Setting one confidence threshold for all fields; totals and part numbers need stricter limits than notes.
- !Skipping review for new suppliers whose layouts the system has not seen.
- !Losing the link between the image and the posted transaction.
- !Not handling multi-page or mixed-document scans.
Frequently asked questions
How accurate is OCR on invoices and packing slips?
It depends on scan quality and layout variety. Clean typed documents extract well, while handwriting and skewed photos need review. Measure on your own documents and tune thresholds per field rather than relying on a headline percentage.
Can the result post to ECI M1 automatically?
It can, through a supported M1 interface, but start with a confirm step. Auto-post only for suppliers and document types where the match rate has been consistently high for a while.
Does M1 have built-in OCR?
We are not aware of a built-in OCR feature across versions. Some resellers offer add-ons, so ask yours. Otherwise OCR sits beside M1 and writes in through its integration points.
Can drawings be read automatically?
Title block data such as part number, revision and material can be read reliably enough to check against the item master. Interpreting full geometry or tolerances is a harder problem and should stay human-reviewed.
Can this run without internet access?
Yes. Open OCR engines and local models run entirely on your servers. That is the usual approach for defense and aerospace suppliers with controlled documents.
Want Document OCR working on your ECI M1?
A free 30 minute assessment: we map the integration path for your version and setup, and where it should run.
Get a free assessment