COSCO SHIPPING and Huawei Co-Build a Global Benchmark for Intelligent Document Recognition
This site uses cookies. By continuing to browse the site you are agreeing to our use of cookies. Read our privacy policy>
Enterprise
Fostering a robust national transportation network and continuously driving down logistics costs stand as core objectives of China's 15th Five-Year Plan for the high-quality development of modern logistics. In 2025, China's total logistics expenditures reached CNY19.5 trillion, representing 13.9% of GDP. This indicates that considerable room for optimization remains compared to the 8% to 9% benchmark observed in developed Western countries. Accelerating the intelligent transformation of document operations is a critical catalyst for unlocking cost savings and efficiency gains—paving the way to achieve the target of lowering the ratio of logistics costs to GDP to 13.5% by 2027.
As a leading comprehensive logistics and supply chain enterprise in China, COSCO SHIPPING runs full-chain businesses spanning maritime shipping, land transportation, multimodal transport, and cross-border supply chain. Throughout its business workflows, a massive volume of logistics documents is generated, including bills of lading, booking notes, packing lists, and customs declarations. As the core medium integrating information, freight, and revenue flows, documents underpin every logistics link, from sorting, consolidation, and loading to transportation and storage. Document processing capability thus directly dictates the overall operational velocity and performance of the entire supply chain.
As COSCO SHIPPING's business scale grew, the company faced increasingly acute bottlenecks in its traditional manual document processing mode.
• Complex document formats across diverse business scenarios: Manually preparing a single customs document, for instance, took between 7 and 20 minutes, and labor costs in document-related roles remained persistently high.
• Error-prone manual information entry: With error rates ranging from 3% to 8%, manual document preparation not only drove up rework but also exposed businesses to compliance risks.
• Insufficient generalization of conventional OCR solutions: They could recognize simple text but lacked the semantic understanding capability. Their accuracy on non-standard documents or those with complex layouts was only around 80%, falling short of the requirements for production and commercial use.
On May 1, 2026, the newly revised Maritime Code of the People's Republic of China took effect, granting electronic bills of lading the same legal status as their paper counterparts. This landmark development provides a legal foundation for digitalizing shipping documents, and as regulations evolve, the industry follows suit. To tackle long-standing document processing challenges in the logistics sector, COSCO SHIPPING, in collaboration with Huawei and its ecosystem partner iSoftStone, has deployed an Intelligent Logistics Document Recognition Solution. By combining Huawei's dedicated document recognition model with the DeepSeek inference model, the solution leverages iSoftStone's Robotic Process Automation (RPA) technology to set up an end-to-end intelligent document processing system. This transforms shipping document operations from paper-based to digital, and from reactive waiting to proactive alerting.
Powered by a specialized multimodal document recognition model coupled with the DeepSeek inference model, the solution is capable of both visual recognition and semantic understanding. By eliminating reliance on the rigid templates typical of conventional OCR solutions, it effortlessly adapts to a vast array of logistics documents and formats, including bills of lading, booking notes, packing lists, and customs declarations. Backed by robust generalization capabilities, it automatically extracts key document fields and generates structured data, overcoming pain points such as inaccurate recognition of complex documents and the inability to interpret complex business semantics.
• Full-process automated transfer: RPA seamlessly processes the structured document data generated by the model, orchestrates complex business workflows, automatically verifies data, flags anomalies, and writes information back into the business systems. This completely replaces tedious, repetitive manual tasks, such as copying, data entry, and verification, realizing a fully automated document lifecycle from initial acceptance to final archiving.
• Built-in business rules for verification: Drawing upon an extensive logistics industry rule library, the solution automatically executes document-by-document and logical verifications to proactively intercept field conflicts and data discrepancies. This preemptive approach drastically reduces rework rates and mitigates compliance risks.
• 24/7 uninterrupted operations: The solution enables centralized processing of a huge volume of documents during peak hours while significantly alleviating the workload on document operations personnel during night shifts and major shopping events.
Since the project went live, the traditional operational mode, which relied heavily on manual document preparation and review, has undergone a complete transformation. The accuracy of complex document recognition has improved dramatically compared to conventional OCR solutions. Meanwhile, the processing time for preparing a single customs document has plummeted from 7 to 20 minutes to just a few minutes. By automating a massive volume of repetitive data entry tasks, the solution has substantially reduced the labor investment in document processing roles. Furthermore, through a dual-verification mechanism combining advanced models with business rules, the document entry error rate has been sharply reduced, minimizing operational risks stemming from rework and discrepancies. Consequently, the freed-up workforce can be redirected toward high-value tasks such as customer service and risk analysis, empowering COSCO SHIPPING to cut costs while boosting efficiency.
This solution was honored by The Chartered Institute of Logistics and Transport (CILT) with the "Golden Key" case award for reducing logistics costs, establishing a replicable model for the digital transformation of global shipping documents.
Looking ahead, COSCO SHIPPING will continue to deepen joint innovation with Huawei and iSoftStone in large model–enabled logistics scenarios, expanding into areas such as electronic waybill parsing within the single-document multimodal transport system and intelligent processing of cross-border documents. By continuously honing its intelligent document recognition capabilities, the company aims to deliver replicable, practical models for the digital upgrade of China's logistics sector, steadily driving down overall logistics costs.