On June 11, 2025, Mercari launched its internal radio-style talk session series “Leaders Talk! AI/LLM Behind the Scenes Open Door.” The guest for the fourth session was Yohei Miyasaka (@yohei), who leads customer service.

In customer service today, a fundamental transformation driven by AI/LLM is underway. This goes beyond reducing costs and improving efficiency—it’s about redesigning the customer experience itself. We took a closer look at what’s happening on the front lines and the future we are aiming for. As in the previous session, the moderator was Gomi Hayakawa (@gomichan) from the AI/LLM Office.

*This article was composed and written using AI tools based on the day’s audio recording. This English version was translated from Japanese using a custom generative AI model interface trained using Mercari style guides, glossaries, and past Mercan articles.

The Reality of Customer Service and the Unchanging Customer Experience

@gomichan: Today’s guest is @yohei, who leads customer service. Thanks for being here!

@yohei: Thank you for having me.

@gomichan: What parts of customer service are you responsible for?

@yohei:  I oversee customer service operations across Marketplace, Fintech, and the group as a whole.

@gomichan: So you’re supporting everyone listening today behind the scenes too. How do you see the evolution of AI tools and technologies in customer service?

@yohei: I first started using generative AI about two or three years ago. Since then, I’ve regularly checked global customer service case studies. My strongest impression was, “Finally, something like this has arrived.” Until then, customer service technology was mostly about internal efficiency, and it hadn’t yet transformed the customer experience itself.

@gomichan: So efficiency and cost-cutting have been the main focus, but the customer experience itself hasn’t changed?

@yohei: I don’t think it has. Even when I contact other services myself, I end up searching FAQs, and if I can’t find a solution, I wait for a response by phone, email, or chat. I think this experience hasn’t really changed in 20 or 30 years.

The Ideal State of Customer Service: No Inquiries at All

@gomichan: On the flip side, what would be the best possible state for customer service?

@yohei: The ideal state is one where inquiries don’t occur in the first place. The very act of needing to contact support already creates stress for the customer.

@gomichan: Not just making the contact points easier to find, but ensuring customers don’t run into issues at all, right?

@yohei: Exactly. Making it harder to contact us for the sake of reducing inquiries isn’t the point. The essence is that the service improves, inquiries remain easy to make, and yet customers no longer feel the need to reach out.

@gomichan: Last year, when we received some tough feedback on social media, how did the customer service team take it?

@yohei: The team members handling operations felt the issues keenly. Honestly, though, we hadn’t been able to implement fundamental improvements. I see that feedback as simply surfacing at that particular timing.

@gomichan: What specific issues had you recognized?

@yohei: Simply put, operations had become too complex. As our services grew, the number of inquiries from customers increased year by year. This created a need for even faster responses in order to maintain service levels.

At the same time, diversification of our service lineup and years of accumulated operational adjustments made customer support more complex than before. As a result, the skill level required of staff increased, and both the number of exchanges per case and the time required to resolve issues (SL*) showed an upward trend.

*SL: Service Level. An indicator that visualizes service standards.

@gomichan: I’ve seen operations on-site, and the time constraints look pretty tight.

@yohei: Yes, we’re expected to handle cases within very limited timeframes.

@gomichan: For example, staff need to look at the customer chat exchanges, the product photos, and the inquiry itself—and still respond within that limited time, right?

@yohei: That’s right. Ensuring customers can use Mercari comfortably is our top priority, and everyone involved wants to solve their problems. Responding quickly to the massive number of inquiries is also crucial for the experience, so we constantly struggle with balancing complex operations and the need for speed.

Understanding Cost per Transaction for More Precise Management

@gomichan: Customer service is often seen as an area where companies try to cut costs. How is Mercari approaching this?

@yohei: It’s true that cost reduction has been a strong focus at Mercari too. But instead of focusing on reducing the total cost itself, we look at “customer service cost per transaction.” For example, if transactions increase by 30% and customer service costs rise by only 20%, that’s a success from a cost perspective. The ideal is to improve quality while driving that percentage even lower.

To repeat what I said earlier, the most important thing is enabling customers to use the service comfortably without needing to contact support in the first place. That requires improving the product itself, so we work very closely with the product team.

The Essence of Automation Is Creating Customer Value

@gomichan: So on one hand, you feed issues from customer service into the product to eliminate customer pain points, and on the other, you handle actual inquiries instantly through AI automation. You’re aiming to streamline both sides, right?

@yohei: Exactly. But when it comes to automation, I think we need to be careful with the nuance. Many people imagine automation simply as taking existing tasks and automating them to cut costs. But that’s not the essence of AI technology.

@gomichan: What do you mean by that?

@yohei: We should be thinking about how to transform value for customers. Instead of focusing only on cost-cutting and operations, we should concentrate on what kind of experience AI can create.

When customers face problems, what they expect most from customer service is a quick resolution.

AI can enable customers to make inquiries anytime, in their own language, and get issues resolved immediately. Overcoming the traditional barriers of “time” and “language” that often limited customer service is the most critical value we aim to provide.

Aiming for Resolution in Five Minutes

@gomichan: I’d like to revisit the vision for the future of customer service. What struck me was your goal of “resolving every issue within five minutes.” Could you elaborate on that?

@yohei: The simplest way to explain it starts with our mission: Why are we here? Our mission is to pursue a customer experience that requires no extra effort—including reducing the need for inquiries. Extending from that is the vision of resolving issues within five minutes.

@gomichan: Honestly, no one really wants to contact support if they can avoid it.

@yohei: Exactly. Our response rate within 24 hours is 95%, and we work hard to reply quickly. But getting to a full resolution can still take 50 hours. Right now, reaching fundamental resolution takes quite a bit of time, especially in cases where two parties are directly involved—it can take even longer.

We see the stress customers feel while waiting as part of the overall burden. That’s why we’re always looking for ways to resolve existing inquiries faster.

@gomichan: Some people in customer service worry that new technology might replace their jobs. How do you see the balance between people and technology?

@yohei: It’s not realistic to replace everything with technology. For complex or emotional cases, human involvement is essential. One unique aspect of Mercari’s customer service is that there are always two customers. In person-to-person transactions, both sides have their own claims and perspectives. In those cases, it’s important to come up with a solution that both parties can accept, and that’s something only people can do.

@gomichan: How do you see customer service evolving from here?

@yohei: @yohei: I see customer service not as a cost center, but as a driver of customer lifetime value (LTV)—the total profit a customer brings to the company over their entire relationship.

When trouble arises, many customers leave without ever contacting support. I want to save every one of them.

@gomichan: What do you see as the ultimate measure of success?

@yohei: Success would ultimately mean reaching zero inquiries. We’ll keep experimenting and improving until we create a service that customers can use comfortably without ever needing to contact us.

@gomichan: Finally, could you share a message about what lies ahead?

@yohei: Our direction is clear. With a positive mindset and curiosity, we’ll keep moving forward. My goal is to build an environment where customers can use Mercari with confidence and enjoyment.