How to add document OCR to QAD Adaptive ERP
Short answer
Capture paper and PDF documents with OCR plus a layout-aware extraction model, score confidence per field, match against QAD purchase orders and receipts, and post through QXtend or web services after review. Invoices, packing slips and certificates of conformance are the practical starting points.
QAD customers in automotive, medical device and discrete manufacturing receive a continuous flow of documents that carry the facts the ERP needs: supplier invoices, packing slips, certificates of conformance and analysis, material test reports, inspection reports and customer drawings. Many arrive as scans or photos from the dock.
Plain OCR returns text but not meaning. The valuable part is turning a document into fields, such as PO number, line quantity, lot, heat number and certificate status, and checking them against what QAD expects. That is a matching problem as much as a reading problem.
This page treats OCR as a capture and validation layer on QAD. It focuses on documents, not email routing, which is covered separately, so you can adopt it for scanner, mobile and portal inputs as well.
What teams use it for
Three-way match for supplier invoices
Extract header and lines, find the PO and receipts in QAD, compare quantity and price, and prepare the supplier invoice for posting with variances shown.
Packing slip to receipt
Read the packing slip at the dock, match to the open PO lines and pre-fill the receipt with quantity, lot and serial where present.
Certificates of conformance and analysis
Extract lot, heat, specification and results, compare against item specs and attach the certificate to the receipt record or inspection.
Material test report checks
Validate chemistry or mechanical values against limits for the item and flag out-of-range values before material is released.
Customer drawing and PO attachment indexing
Identify drawing number and revision on incoming drawings, compare to the item revision in QAD and flag mismatches.
Customs and shipping documents
Extract commercial invoice, HS code and country-of-origin data to support import receipts and landed cost entries.
QAD Adaptive ERP integration surface
- Open PO and receipt data from QAD tables such as po_mstr, pod_det and receipt history, read through SQL-92 or QAD services, for matching extracted values.
- Item master and quality specifications (pt_mstr and related quality or inspection tables, depending on whether you use QAD Quality or custom fields) for certificate and test report validation.
- QXtend or QAD web services to post receipts, update lot attributes, and create supplier invoices through standard programs.
- Document linkage: attach files in a document management system or a file share, and store a reference or URL on the QAD record. Native attachment capabilities vary by release and configuration.
- Capture inputs: network scanners, MFP email-to-folder, mobile photo upload from the dock, and a supplier portal. All feed the same pipeline.
- Master data quality is the true integration surface: supplier part numbers, item cross references, units of measure and tolerance settings determine match rates.
How it works
- 1
Choose document classes
Start with two. Invoices and packing slips are common. Define the fields you need for each and which QAD record they must match.
- 2
Build the capture inputs
Set scan folders, a mobile upload page and a mailbox. Standardise image quality: 300 dpi, deskewed, with multi-page handling.
- 3
OCR and structure extraction
Use OCR plus a layout-aware or vision-language model to produce fields and tables. Store the page image, text, fields and per-field confidence.
- 4
Match against QAD
Look up the PO and lines, apply supplier part cross references, convert units, and compute variances. A document with an unmatched PO goes to exceptions.
- 5
Apply confidence and business thresholds
Set thresholds per field. High-confidence and within-tolerance documents proceed to a draft. Anything below goes to a person, with the failing field highlighted on the image.
- 6
Post through QAD logic
Create the receipt or invoice draft via QXtend or services. Keep the document image linked to the resulting transaction.
- 7
Learn from corrections
Store reviewer fixes per supplier template and feed them into rules and evaluation sets. Track field accuracy over time.
Example flows
Invoice with a quantity variance
Trigger: A supplier invoice PDF is scanned by AP.
- OCR and extraction read supplier, invoice number, PO, lines and totals.
- The pipeline fetches PO lines and receipts from QAD.
- Line 3 was invoiced for 100 but received 90; variance is flagged.
- AP sees the image with the line highlighted and chooses to hold or short-pay.
Result: The mismatch is caught before posting, with the evidence on screen.
Certificate of conformance at receiving
Trigger: A receiving clerk photographs a CoC with the packing slip.
- The mobile upload sends images to the pipeline.
- It extracts lot, heat and specification, and finds the open PO line.
- It compares values with the item's required specification.
- Compliant documents attach to the receipt; others go to quality review.
Result: Quality gets complete, searchable records without manual indexing.
Drawing revision mismatch
Trigger: A customer sends a drawing with a PO.
- The model reads drawing number and revision from the title block.
- It compares against the item's current revision in QAD.
- A mismatch creates a task for engineering before the order is released.
Result: Revision errors are caught at order entry instead of on the shop floor.
Keeping data on-prem
Documents are often the most sensitive data you hold: pricing, drawings and test results, sometimes export-controlled. Run OCR and the extraction model on your own servers with open-weight vision-language models and standard OCR engines, store images in your own repository, and call QAD from inside the network. There is no need for any document to leave your boundary. Plan GPU capacity for peak scan times, queue large batches overnight and keep a fallback to manual entry if the service is down.
Security checklist
- Encrypt document storage and restrict access by document class, since invoices and test reports have different audiences.
- Do not send images to public APIs if they include controlled technical data or customer confidential pricing.
- Keep original image, extracted fields and reviewer changes together as an audit record.
- Block auto-posting when bank details, remit-to addresses or totals differ from supplier master history.
- Apply retention rules to images and extracted text in line with finance and quality obligations.
Ways to get it done
Commercial AP or capture product with a QAD connector
Effort: 6 to 12 weeks including tuning
Invoices are the main need and a vendor supports QAD (check your version and hosting model).
Open-source OCR with custom extraction and QXtend
Effort: 2 to 4 months for two classes
You want broader document types and control, and can maintain models and rules.
Partner build with Netray
Effort: Scoped after discovery
You want private, on-prem extraction with QAD matching, review screens and confidence rules, designed with your QAD team.
Pitfalls
- !Judging OCR by character accuracy instead of field accuracy and match rate against QAD.
- !Weak supplier part cross references, which causes unmatched lines even when extraction is perfect.
- !Ignoring multi-page and multi-PO invoices, which break simple header-line assumptions.
- !No threshold governance, so a model update quietly changes straight-through rates.
- !Photos from phones with glare or skew; set capture guidelines and reject poor images at upload.
Frequently asked questions
Does QAD Adaptive ERP include OCR?
QAD provides document and supplier invoice capabilities in some editions and through partners, but coverage depends on your version and configuration. Check what you have. If it covers your document types and accuracy needs, use it. Add a separate OCR layer when you need certificates, packing slips or private on-prem extraction.
How do confidence thresholds work?
Each extracted field carries a confidence score and each business rule a tolerance. Documents where every critical field is above threshold and matches QAD proceed to a draft. Others go to review with the weak fields highlighted. You tune thresholds from measured correction rates.
Can OCR post receipts directly into QAD?
It can create receipts through QXtend or services, but most teams start with drafts for human confirmation. Direct posting can be enabled for narrow cases, such as exact quantity matches from trusted suppliers, once the pipeline has a track record.
What about handwritten notes and poor scans?
They are the weak spot. Modern vision models handle some handwriting, but accuracy is lower and the cost of a mistake is higher. Route low-confidence pages to review and improve capture quality first, since better scans give larger gains than a bigger model.
Can we keep all documents on our own servers?
Yes. OCR engines and open-weight vision-language models run on-prem. You provide GPU capacity and storage. That is the typical architecture for defense, aerospace and medical customers that cannot send supplier documents to a public AI service.
Want Document OCR working on your QAD Adaptive ERP?
A free 30 minute assessment: we map the integration path for your version and setup, and where it should run.
Get a free assessment