440 Meridian Ave, San Jose, CA 95126 +1 (408) 555-0194 hello@brightlooplabs.com

HomeProjectsDocuMind Pipeline

DocuMind Extraction Pipeline

Document AI for ParcelPost Logistics — 40,000 carrier invoices a month captured, validated and posted with no manual keying.

  • Document AI
  • Client: ParcelPost Logistics
  • Timeline: 9 weeks
  • Live since 2024
Reviewing an invoice for automated data extraction

Overview & Challenge

Eleven Clerks Vs. Forty Thousand Invoices

ParcelPost receives carrier invoices in every imaginable shape: PDFs, phone photos, scans with coffee stains, 400+ layouts across 60 carriers. Eleven clerks typed them into the TMS — slowly, with errors that sparked endless carrier disputes.

Off-the-shelf OCR had been tried and abandoned: it read the text but couldn't understand which number was the fuel surcharge and which was the linehaul.

Validating extraction accuracy with the finance team High-throughput invoice processing station

Approach & Solution

Layout-Aware, Confidence-Gated

We trained layout-aware extractors on 8,000 annotated invoices, then wrapped them in business rules: totals must reconcile, surcharges must match rate cards, duplicates are rejected. Only clean postings reach the TMS.

  • 400+ layout variantsNew carrier formats onboarded in days, not projects.
  • Three-way validationInvoice vs. rate card vs. delivery record before posting.
  • Two-minute review queueClerks now check exceptions instead of typing.
99.1%
Field-level extraction accuracy
40k
Invoices processed monthly
−81%
Cost per invoice processed
−64%
Carrier billing disputes

Related Service

Built with our Image Processing practice — OCR, layout parsing and validation.

Get Updated With The Latest Newsletter