Q1. AI adoption is accelerating in logistics, but translating that into changes in day-to-day work remains difficult.
What do you see as the biggest obstacle?

I think the biggest obstacle is that the conditions needed to put AI to work aren’t fully in place yet. Logistics depends on shippers, freight forwarders, ocean carriers, ports, terminals, and inland transportation providers working together to keep a supply chain moving. Each uses different systems, and information is scattered across them. That makes it difficult to connect the necessary data in the first place, then use AI’s results within existing workflows.

A Gartner survey released in April found the same problem. Between October and November 2025, Gartner surveyed 140 chief supply chain officers at companies with annual revenue of at least $250 million. Of those surveyed, 56% identified integrating AI into existing systems and processes as a major challenge. Making AI part of everyday work has become a bigger challenge than generating interest in it or demonstrating its technical potential.

This matters especially in logistics. Take an estimated time of arrival, or ETA. The person managing a shipment needs more than an updated prediction. They have to use that information to decide when to schedule receiving, when to have warehouse capacity and trucks ready, and which arrival estimate to use when giving the customer a delivery date. An AI result becomes useful on the job when it informs those decisions.

That was our focus with AI-ETA, which Tradlinx launched this year. A conventional carrier ETA gives you a single arrival time. AI-ETA provides a range of possible arrival times, from early arrival through a baseline estimate to delayed arrival, along with their probabilities. Teams can choose the estimate that fits the task instead of building every plan around one date.

For example, they can plan production and receiving around the baseline arrival estimate, while using the probability of arrival within 24 hours to arrange warehouse or yard slots. For decisions that call for more caution, such as delivery commitments, they can account for the possibility of a delay. Freight forwarders can check shipments with increased delay risk first and respond before customers start asking questions.

We also tested prediction performance against actual arrival data. In a comparison covering 508,055 records with confirmed arrivals through March 2026, AI-ETA reduced average prediction error by 31.4% compared with carrier ETAs. As arrival approached—the stage when teams begin preparing warehouse capacity, trucks, and labor—92.5% of actual arrivals fell within 24 hours before or after the predicted time.

What will make logistics AI competitive is how well it connects the information it produces to decisions and actions on the job. The number of features will matter less. When people can use AI’s results to prepare earlier and adjust plans faster, we’ll begin to see changes that they can feel in their daily work.

Q2. Tradlinx is designing its AI to explain the basis for its predictions. Why is that important?

In logistics, a prediction feeds directly into an operating plan. If AI predicts that a shipment will arrive several days late, the person handling it has to revisit receiving and truck dispatch schedules, as well as the delivery date promised to the customer. If all they see is the expected number of days of delay, it’s hard to trust that result enough to change several plans right away. They need to know which factors the prediction takes into account: conditions at the port, changes in vessel operations, or delays accumulated on earlier legs of the journey. That’s how they judge whether the result makes sense.

This is why we place so much importance on HPI-AI, or Human-Perspective Interpretable AI. It’s intended to make AI predictions understandable from the perspective of the people working in logistics. To use a prediction in a decision, they need to know more than which information went into it. They also need to understand how that information relates to the criteria they use in their work. HPI-AI aims to reflect how they make decisions and clearly show how the relevant factors relate to the predicted outcome.

For example, if port operating conditions are identified as a major factor in a predicted delay, the person managing the shipment can examine exactly what information was used and how it relates to the result. They can then assess whether the prediction is reasonable and decide whether to use an alternative route or adjust the delivery schedule. If they decide to change the plan, the evidence provided by the AI also gives them a basis for explaining that decision to colleagues, customers, and partners.

We want people to interpret the evidence in the context of their own work and decide for themselves whether the result is trustworthy. They shouldn’t have to accept an AI explanation at face value. That evaluation is what allows an AI result to lead to an operational response.

We’re continuing the research and development needed to put this approach into practice. Tradlinx was recently selected for the Scale-up TIPS R&D program, overseen by the Ministry of SMEs and Startups and the Korea Technology and Information Promotion Agency for SMEs. We’re conducting the research with Pusan National University and the Korea Advanced Institute of Science and Technology, or KAIST. We’re developing a multi-agent system in which specialized AI agents work together on arrival prediction, early delay detection, and multimodal route recommendations. We plan to incorporate HPI-AI-based decision support so logistics professionals can interpret and use the evidence behind those predictions and recommendations, using the same criteria they already use to make decisions.

We want to build AI that helps people review its analysis against what they know from the field and the requirements of their work, then decide how to respond. We aren’t trying to have AI make every judgment for them. When it supports that process, AI can become a partner in decision-making instead of another thing someone has to check.

Q3. Tradlinx’s international customer and partner base is growing rapidly. What is your global strategy for sustaining that growth?

The clearest lesson from entering international markets is that the technology and experience we’ve built over more than a decade in Korean logistics also hold up overseas. Our customer base is growing quickly across a range of countries, and we’re seeing more new customers come through referrals from existing customers. Recently, more local companies and resellers in several countries have also been approaching us about working together. Experience with our products is bringing us new customers and partners. That goes beyond attracting interest overseas and gives us reason to believe our products and technology can compete globally.

We’re expanding the ways we reach customers to support that growth. Along with strengthening our global sales team at headquarters, we’re hiring locally in regions where demand is increasing. We plan to gradually expand our local teams and partner network as we assess market response and business opportunities.

As we engage with customers and partners in more places and through more channels, we gather a wider range of information to improve our products. We identify needs and areas for improvement that recur across markets, then use those findings to set our product development direction and priorities. International expansion is giving us a foundation for advancing our products and technology as we grow our customer base.

We’ll continue expanding where and how we reach customers, while improving our ability to respond quickly to local needs through product and technology development. Through this approach, we will strengthen our competitiveness in each market so that customers choose our products and continue to use them. We will grow into a supply chain technology company trusted by businesses around the world.