The DX Gap in the Supply Chain

Across industries, the importance of DX continues to grow as technologies such as generative AI evolve at a rapid pace. At the same time, the gap between industries that are advancing in digitalization and those that are not has been steadily widening. This disparity is particularly evident in supply chain operations, where complex, long-established business practices remain deeply embedded. Processes reliant on fax transmissions and handwritten documents are still common, creating not only inefficiencies but also risks such as human error and excessive dependence on individual expertise—factors that can ultimately undermine overall industrial competitiveness.

The business domains of a general trading company like Marubeni are closely linked to these challenges. At the same time, this positions Marubeni to play a role in advancing DX at the operational frontier. An AI-powered OCR solution jointly developed by Marubeni Logistics Corporation and Marubeni’s Digital Innovation Department (DI Department), and introduced at a logistics hub serving a major snack manufacturer, represents a concrete step toward addressing this long-standing situation.

Marubeni Logistics handles order processing and logistics operations for a rice cracker manufacturer operating multiple production and distribution sites across Niigata Prefecture. In the confectionery manufacturing industry, fax-based ordering remains widespread, with approximately 30% of orders processed by Marubeni Logistics still received via fax. While several attempts had previously been made to improve efficiency using OCR, differences in order form formats across customers and the prevalence of handwritten entries posed significant challenges to accuracy, preventing full-scale deployment. As a result, staff were required to manually review and input each order individually—a process that required considerable time and effort.

Semi-Automating Data Entry Through Generative AI

To address this long-standing issue, Marubeni Logistics partnered with the DI Department, which leads the development of a wide range of digital technologies within the Marubeni Group. A key turning point was the rapid advancement of generative AI. The project began with preliminary validation in the summer of 2023, when generative AI gained broader attention, before moving into full-scale consideration for operational deployment the following year.

During system development, multiple approaches were explored through repeated proof-of-concept testing at operational sites. The focus extended beyond simple text recognition to include detailed prompt engineering tailored to actual workflows—for example, enabling handwritten order codes to be interpreted correctly or preventing outdated delivery dates from being displayed. Through continuous refinement informed by on-site feedback, accuracy was gradually improved to a level suitable for practical use.

As a result, a new workflow was established to replace the conventional three-step process of sorting, manual data entry, and content verification based on faxed order forms. Under the new system, order forms are converted into PDFs, processed through OCR, and structured Excel data is generated using generative AI. After human review and correction, order data can be batch-imported into the system. Whereas multiple staff members were previously required, human involvement is now largely limited to verification and correction, significantly reducing the likelihood of errors.

These improvements are also reflected in measurable results. Order processing tasks that previously required a cumulative 17 hours per day have been reduced to 9 hours following the system’s introduction, achieving an eight-hour reduction in workload. Given the strict daily deadlines associated with order processing, staffing demands for handling fax-based orders had long posed a challenge, and the solution has also contributed to reducing labor costs.

While achieving 100% accuracy was not feasible at the initial stage—given the need to balance accuracy, speed, and cost—continuous improvements have steadily enhanced system performance. The burden of human verification and correction has also been reduced, bringing the process closer to semi-automation.

Expanding Applications Within and Beyond the Marubeni Group

Today, the system is poised to extend beyond Marubeni Logistics’ logistics centers. Building on the tool originally developed for order forms, plans are underway to create a more general-purpose solution capable of handling a wider range of documents, including invoices. Looking ahead, integration with “Marubeni Chatbot,” a generative AI chatbot provided internally within the Marubeni Group, is also under consideration, with the aim of making the solution accessible to a broader range of employees. In parallel, Marubeni Logistics is deepening its collaboration with the DI Department by advancing development in additional areas, such as document matching for trade operations.

Such flexible and rapid development and deployment are made possible by the ability to complete the entire process—from planning and development through to implementation—within the Marubeni Group. Close collaboration between business units with a deep understanding of on-site challenges and specialized teams with advanced technical expertise serves as a key driver of Marubeni’s DX initiatives.

While engaging external vendors can enable rapid service development, it can be difficult to foster commitment that extends beyond conventional client–vendor relationships. In contrast, the mission of the DI Department goes beyond addressing individual business unit challenges; it also involves accumulating the insights and know-how gained through these initiatives as shared assets for the entire group, thereby creating new value. By engaging directly with real-world operational challenges and mobilizing the collective strengths of the group, DX initiatives of this kind can be expected to contribute to meaningful transformation across a wide range of industries.