Skip to content
AMTEXConsulting

OCI Document Understanding and intelligent document processing

Invoices, POs and forms that read themselves into Fusion.

OCI Document Understanding, custom extraction models and Fusion Intelligent Document Recognition, wired through OIC into Payables, Receivables, HCM and Procurement with confidence thresholds and a review queue.

Document pipeline into Fusion
  1. Capture

    • Email
    • Supplier portal
    • SFTP / scans
  2. Understand

    OCI Document Understanding

    Classify · extract · custom models

  3. Validate & review

    • Master data & PO match
    • Review queue (OPA / APEX)
  4. Post

    • OIC
    • Fusion Payables / HCM / Procurement
    • Object Storage archive

Overview

Document Understanding, done properly.

Accounts payable teams still key invoices. HR still reads certificates. Procurement still reconciles paper acknowledgements. The documents arrive by email, portal and SFTP in every layout. Document intelligence turns them into structured records, and the difference between a demo and a working system is what happens to the twenty percent the model is unsure about.

AMTEX builds end-to-end document pipelines on OCI: capture, classify, extract, validate against Fusion master data, route low-confidence fields to a reviewer, then create the transaction through OIC. We also implement and tune Fusion's own Intelligent Document Recognition for Payables where it fits.

Technologies

  • OCI Document Understanding
  • Fusion Intelligent Document Recognition
  • Oracle Integration Cloud
  • OCI Object Storage
  • Oracle Process Automation
  • Oracle APEX
  • Document Intelligence

What we deliver

Capabilities

Everything AMTEX builds, configures and runs across Document Understanding.

  • OCI Document Understanding

    Pre-trained invoice, receipt and passport models, key-value and table extraction, document classification.

  • Custom extraction models

    Trained on your layouts for the documents pre-trained models miss: delivery notes, certificates, contracts, forms.

  • Fusion IDR for Payables

    Implementation and tuning of Intelligent Document Recognition, supplier-specific rules, and the interface to invoice workflow.

  • Validation against master data

    Supplier, PO, tax and GL matching before anything is created, so the extraction is trusted, not just produced.

  • Confidence thresholds & review

    Field-level confidence, a review queue in OPA or a lightweight APEX app, and feedback that improves the model.

  • Capture channels

    Email mailboxes, supplier portal uploads, SFTP drops and scanned batches into one Object Storage landing zone.

  • OIC orchestration

    Classification to extraction to validation to Fusion REST or FBDI, with tracking by document id.

  • Contracts & HR documents

    Clause extraction, expiry tracking and right-to-work or certification capture into HCM Document Records.

  • Audit & retention

    Original images retained with the transaction, every extraction and correction logged.

The hard parts

Where these projects become difficult

The places programmes slip, and what we do about each one.

  1. 01

    The long tail of layouts

    The problem

    The top twenty suppliers extract cleanly; the other four hundred do not.

    How AMTEX approaches it

    Pre-trained models for the common cases, custom models for the volume outliers, and a review queue sized honestly for the rest.

  2. 02

    Extraction without validation

    The problem

    A perfectly extracted invoice for a supplier site that is inactive still fails in Fusion.

    How AMTEX approaches it

    Master-data validation and PO matching as pipeline steps, with the same business-fault handling we use in every OIC estate.

  3. 03

    Reviewers ignored

    The problem

    The review queue fills up and the process is slower than keying.

    How AMTEX approaches it

    Threshold tuning from real data, review UI designed with AP, and metrics on straight-through rate from day one.

How we work

From first conversation to steady state

  1. 01

    Sample

    A representative document set and the straight-through target.

  2. 02

    Prototype

    Extraction accuracy measured per field on the sample.

  3. 03

    Build

    Capture, validation, review queue and Fusion integration.

  4. 04

    Tune

    Thresholds, custom models and supplier rules improved from production feedback.

FAQ

Common questions

Still keying documents into Fusion?

Send us a sample set. We will measure extraction accuracy on your real documents before anyone commits.