Logistics and Customs Document OCR Platform

The client is a customs brokerage firm. Every day it receives many sets of import and export documents as PDFs: commercial invoices, packing lists, bills of lading, and permits. Many are scans or faxes, with faded and skewed text.

Staff had to retype figures, compare every document, look up HS codes, and then file declarations by hand. This was slow. A single mistyped number could go straight into a customs declaration.

We built an end-to-end document processing platform with eight core capabilities:

• Page classification: splits a combined PDF into individual document types, even when the pages are out of order.
• Two-layer extraction: OCR reads text and tables, while AI reads the page images directly. The two readers run in parallel and cross-verify each other's results.
• Document cross-checking: compares quantities, weights, and values across documents. Any file with a discrepancy is put on hold for a person to review.
• HS code suggestions: AI suggests HS codes based on the goods description and tariff data.
• Customs value calculation: automatically adds up the components of the customs value and converts foreign currencies at the exchange rate.
• Review screen: staff view the results on the web, correct what needs correcting, and approve.
• AI assistant: answers questions about HS codes and customs procedures.
• Filing robot: automatically submits the customs declaration once the file is approved. If a task fails partway through, the robot recovers automatically and continues.

3K TechOne developed nearly the entire system.

• Less manual data entry from documents.
• The process runs automatically from document upload to customs declaration filing.
• Discrepancies between documents are caught before filing, not after.
• The system can read even low-quality scans and faxes.