Logistics AI Needs More Than Optimization: How YUKAI Is Pursuing Both Profitability and Trust | ONZALINX

How ONZALINX, the Company Behind YUKAI, Envisions Logistics Collaboration Across Company Boundaries

AI can create more efficient plans. But companies must also be able to adopt those plans responsibly.

ONZALINX, the company behind the AI logistics platform YUKAI, is working to connect these two goals. For a company that designs and implements logistics systems, “trustworthy AI” is not merely an abstract principle. It is a concrete business challenge: how to use AI in real-world customer operations while enabling companies to collaborate with one another.

Q1. What are your company’s strengths, and what are you currently working on—or planning for the future—in AI?

Our strength lies in understanding the challenges of logistics operations and supporting customers from business process design through system implementation and post-deployment improvement.

Through our warehouse management system, INTER-STOCK, we provide an environment in which customers can continue developing systems suited to their own operations, including through source-code disclosure and support for low-code development and generative AI.

In AI logistics, our earlier initiative, formerly known as “輸快通快”, has now been integrated into YUKAI. YUKAI is a platform that uses demand forecasting and optimization to connect ordering quantities, inventory allocation, and delivery instructions.

What matters to us is not prediction accuracy alone. Our goal is to deliver the products that are needed while reducing capital tied up in inventory and lowering operating costs—turning logistics into a source of profitability.

Q2. Through these initiatives, what challenges do you hope to solve for customers, local communities, and society, and what value do you want to create?

“We want to reduce inventory, but we cannot afford more stockouts.”

We want to help logistics operators address challenges like this not only through experience and intuition, but through data.

As an entry point, before a customer adopts YUKAI, we offer a service that uses the customer’s own sales, purchasing, and inventory data to estimate the potential for profit improvement. The purpose is not to adopt AI simply because it is AI, but to allow each company to determine what value it could actually create for its own business.

These estimates do not guarantee future results, but they provide concrete information that companies can use when considering adoption.

At the same time, we want to broaden access to digital technologies for the small and medium-sized businesses that support logistics. Connecting improvements in individual companies’ management to greater efficiency across the logistics system as a whole—and to lower environmental impact—is part of the social value we aim to create through our business.

Q3. What measures and verification mechanisms have you put in place to enable AI to be used and provided with confidence? Please give a specific example.

One of the things we are working on is linking AI recommendations to grounds for determining whether those recommendations can be adopted.

For example, in a pharmaceutical cold-chain validation prototype jointly developed with the GhostDrift Mathematical Institute, evidence such as temperature records, seals, and custody transfer records was tied to each individual handover decision.

Evaluation proceeds to the next stage only when the required conditions are satisfied. If evidence is missing or the conditions are not met, the process does not advance.

This was a demonstration conducted under defined operational assumptions and does not guarantee the safety of an entire real-world transport operation. We have published the code and reproduction procedures so that the decision can be checked using the same inputs, rules, and implementation.

What matters to us is showing not only what was verified, but also whether the basis for that decision can be checked afterward.

Q4. Based on your experience, what lessons or challenges would you like to share with other companies and industries as we work to expand the “safe, secure, and trustworthy AI” envisioned by the Hiroshima AI Process?

The key point is that being able to optimize something is not the same as being able to adopt that decision as a company.

Even if YUKAI can present logistics options, real-world operations still need answers to questions such as: Was the information used valid? Were the necessary conditions checked? Who approved the decision?

These are issues that AI performance alone cannot solve. They require careful design on the business and operational side.

To put the principles of the Hiroshima AI Process into practice, we believe it is important to move beyond simply stating that “we prioritize safety” and toward mechanisms that verify conditions before a decision is adopted and allow the basis for that decision to be checked afterward.

At HAAC, we hope to bring together these kinds of operational challenges and implementation insights so they can be examined across companies and industries.

Q5. Looking ahead, what kinds of companies and organizations would you like to work with, and what would you like to achieve together by leveraging your strengths?

We would like to combine the capabilities of shippers and logistics companies facing challenges in collaborative delivery and the shared use of vehicles and warehouses with those of technology companies and research institutions.

In collaborative logistics, even if total transportation costs fall, companies cannot participate if delays or operational burdens become concentrated on a single participant. There are also differences between existing systems, as well as legitimate concerns about sharing business data with other companies.

To address these challenges, we announced our concept of a “Verifiable Physical Internet.”

The concept aims to allow each company to retain its own objectives and systems, avoid disclosing more data than necessary, and verify whether the conditions required for joint execution—such as delivery deadlines and acceptable burden limits—are satisfied.

From our position as a company responsible for the design and implementation of real-world logistics operations, we want to make concrete a form of logistics in which companies can collaborate while preserving their differences, without requiring every participant to operate in the same way.

Interviewee

Masaya Higashi
Representative Director
ONZALINX Co., Ltd.