In October 2024, Mercari announced its vision to become a company of AI-led growth, putting technological innovation powered by AI/LLMs at the heart of its business. A little over one year later, AI adoption is steadily advancing and has now become a major movement sweeping through the whole company.

In this time, everything at Mercari has changed—not just our technology, but also the way we think as an organization, how we work, and even our approach to our products. We’re stepping into an exciting new phase where our concrete ideas are moving into action, as top-down and bottom-up approaches come together and employees specialized in all different areas dive into using AI.

How far has Mercari progressed with AI implementation in the year since the last discussion? And what exactly is the AI-Native organization we’re striving to become? We spoke with Masato Yamamoto (@mark), Executive Officer and Senior Vice President of Japan Business; Shunya Kimura (@kimuras), Vice President and Chief Technology Officer of Japan Business; and Gomi Hayakawa (@gomichan), who has been leading the charge on generative AI and now works on the AI Strategy Team, about Mercari’s journey so far and their vision for the future.

Taking generative AI from just another feature to something everyone uses every day

@gomichan: It’s been about a year since the article in October 2024. Looking back on the past year, what changes have you seen?

@mark: It feels like all sorts of things, both big and small, have changed. This past year has brought some major changes, including advances in AI technology. To make that happen, not only did we shift our organizational mindset, but we also updated the company roadmap. It feels like major pieces of this field have moved around.

@kimuras: When we spoke back then, AI was just starting to gain momentum, and I think we didn’t yet have a concrete picture of how we would use it. But in 2025, things really started to take shape—the details have been nailed down, and now our plan for rolling it out across the company is ready to put into action. I think the biggest change is that the excitement, which hadn’t really caught on company-wide before, is now being driven from both the top down and the bottom up.

@gomichan: Honestly, at the end of 2024, we were all wondering what would happen next. But once the new year rolled around, everything kicked into high gear.

@kimuras: Once the new year hit, more and more people started to see how they personally could use AI. That’s why we’ve seen a surge in posts on Slack and X—both inside and outside the company—of people sharing their experiences and spreading new information. A mix of environmental and internal factors came together, and the excitement just kept building.

@gomichan: Actually, this isn’t the first time we’ve seen AI trending within the company. I noticed that there are still some custom emoji in our company Slack workspace that were added back during previous AI trends. Some of our long-time members, having experienced previous spikes in AI’s popularity, might have wondered if this latest surge of excitement was just another passing fad.
However, things are clearly taking a different turn this time around. How is this time different from previous AI trends? What are the major changes?

@kimuras: The first real spike in popularity happened in the latter half of 2017. Everyone was talking about deep learning, but while we kept up the momentum in some areas like AI listing and Trust and Safety’s detection of items that violate our terms of service, these efforts only made up a small part of what was actually happening throughout the company.

Back then, it was tough to make the most of AI unless you had specialized know-how—like that of a machine learning engineer. Having that expertise was necessary to fully leverage AI. In that sense, it was difficult to get everyone across the company on board.

@mark: This past year, both our approach and what we’re focusing on have changed dramatically. In the past, we were proactive about bringing AI into our product—think AI credit and AI listing, for example. But what’s happening now is a whole different story. It’s become something that belongs to everyone—employees, users, you name it. I think the big difference is that there’s a real wave of excitement sweeping through the whole company.

@gomichan: Up until now, it was simply one technology working behind the scenes to support the product. That’s important, of course, but it was only really meant for our users—it wasn’t about how we work. This time, we’re talking about generative AI that we can interact with in natural language, so the experience is totally different—not just for our users, but for the people making Mercari, too. That’s exactly why we’re seeing such a huge wave of excitement right now.

The drive from the top and the passion from the bottom

@gomichan: I think things really started to shift around February 2025. With new models like ChatGPT’s Deep Research and Claude 3.7 Sonnet coming out, we moved from the days of “AI’s convenient, but it makes things up” to a sense of “hey, this might actually be useful!” It was a major turning point when technological changes to AI finally enabled it to clearly keep up with our work.

@gomichan: We distributed Cursor licenses to everyone who wanted one back in March—how did you feel at that time?

@kimuras: What we were worried about in 2024 was, of course, hallucinations. People were saying it wasn’t reliable for important decisions, that there was a risk of mistakes in key summaries, and that it even caused issues in coding. But I think everyone got the hang of it as they kept using it. LLMs still sometimes hallucinate, but they’ve become much more convenient to use. People are getting used to leveraging them for business in a way that minimizes the risks of hallucinations.
Back in March, when I suggested that everyone start using Cursor, I was a little worried about whether people would actually give it a try. A lot of people still didn’t trust AI, and there was definitely some risk in bringing it into our workflow. But once we set a goal to go all-in and use it for everything, and handed out licenses to everyone, we were pleasantly surprised by how many people really started using it.

@gomichan: I didn’t see that coming, either. Before that, I remember a lot of engineers saying things like, “Wouldn’t it be faster if I just wrote the code myself?” But once we actually got started, it turned out to be useful for a variety of tasks, and I was honestly impressed by how well it was working.

@mark: This was truly surprising. It wasn’t just that the top-down approach worked—the bottom-up energy was incredibly strong, too.

At this moment, everyone—from engineers and engineering managers to VPs—was on board and ready to dive in. Everything aligned perfectly for us to start implementing it. Thanks to all the groundwork we’d done and the technology just happening to come together at the right moment, nearly everyone jumped in and started using it at once. What really amazed me was seeing all sorts of hackathons organized by different members.

@gomichan: Spontaneous study groups were happening everywhere, right?

@mark: Back then, people were voluntarily organizing week-long hackathons where they used nothing but Cursor. The learning that came out of that was truly immense. You can’t make the most of AI unless you understand the challenges involved, so I was genuinely delighted to see bottom-up initiatives taking off.

It may look like everyone suddenly jumped in, but in reality, it wasn’t such a disruptive change. Since October 2024, I’ve been steadily sharing with not just engineers, but also people in sales, customer service, and other areas what AI is all about and how it can be used. As questions from other teams kept building up, I think both the bottom-up and top-down approaches came together at the same time, creating real momentum.

Several years of commitment to ethics and a strong foundation became our greatest strengths

@gomichan: People tend to focus on how AI is actually being leveraged, but I think the preparation that happened beforehand was just as important—setting up Security Team guidelines, working on AI ethics, organizing data, and more. Looking back, @kimuras, what kind of groundwork do you remember putting in place?

@kimuras: We’d been having plenty of debates about AI ethics long before generative AI came onto the scene. Once it did, conversations about risks really heated up around the world—but at Mercari, we actually started addressing these issues about four years ago.

There have been global discussions about AI unintentionally promoting discrimination, and concerns over the fairness of machine learning outcomes. That’s why at Mercari, we’ve been having ongoing discussions about how to use AI responsibly—making sure we don’t infringe on anyone’s privacy—and how to build a system that can quickly spot any risks if they pop up. In doing so, we developed theories and created guidelines. Thanks to this foundation, lively discussions about the risks and benefits of AI are now taking place in open-door sessions (internal events).

If we hadn’t laid the groundwork and just suddenly announced, “Hey, let’s have an open-door session about the risks of AI!” I doubt anyone would show up. I’m grateful to the team members who keep at it, never giving up and always looking ahead to the future.

The same goes for building our data infrastructure. Thanks to the solid foundation we spent years building, tools like Socrates—which is driving AI utilization on the business side of things—have suddenly become widely used across the company.

At Mercari, we’ve been promoting the use of AI ever since the days when machine learning started trending. I think that those years of ongoing efforts helped turn this momentum into a wave that’s swept through the entire company.

ROI is a trap—redesigning organizations to shift the focus from people to AI

@gomichan: So, what kinds of changes do we expect this wave to bring?

@mark: Right now, we’re tackling AI as a big topic under the banner of “AI-Native.” This approach is about treating AI as more than just a surface-level tool—instead, seeing what happens when you think of AI as a foundation. If you just try to use it for a quick fix, you’ll find it doesn’t really mesh with the way things are done now. I think it’s important to take an approach that turns conventional thinking on its head.

To create a significant shift, the change needs to happen across the board, not just in isolated areas. We’re working to reinvent our three major pillars—our product, our people, and our organization—by making them AI-Native.

@kimuras: Thanks to our various initiatives, we’re already seeing some impressive results in the numbers. For example, as of July 2025, 95% of our employees were using some kind of AI tool, and our development speed had jumped by 64% compared to last year. But we don’t see this as the finish line—in other words, we’re not calling ourselves truly AI-Native just yet. In fact, we believe this is where the real transformation truly begins.

@gomichan: So it’s only just beginning? What exactly do you mean by “real transformation”?

@kimuras: “Real transformation” means a fundamental shift in the way we think. Up until now, work and organizations have always been designed around people—built on the assumption of what humans can handle and process.

But in a world where AI is the foundation, everything starts from a different place. Now, we can design our organizations and ways of working by starting with the vision we want to achieve and the value we want to deliver to our users, and working backward from there.

In other words, instead of just plugging AI into our existing workflows, we start by asking what we want to achieve and then optimize the systems, teams, and decision-making processes from the ground up to make it happen. I think this is what being truly AI-Native is all about—the real vision we’re aiming for.

@gomichan: By keeping our sights set on the ideal we want to achieve, we’ll keep driving change forward. On the flip side, is there anything you’re wary of?

@mark: I want to be thoughtful about making this wave even bigger. Waves cancel each other out when they meet coming from opposite directions. So, if you want to keep the momentum going strong, it’s important to avoid creating any negative or counter waves.

In the short term, this means making sure the business doesn’t need to set aside a budget for AI initiatives, and clearly separating product priorities from those of AI implementation. We absolutely have to implement AI, but we need to make sure it doesn’t have any negative impact on our existing businesses. It’s crucial to set up a situation where we don’t even have to question whether to do it or not—this helps prevent any negative waves from forming.

@kimuras: ROI is such a trap, you know? It’s not about deciding whether or not to do it in the short term—doing it is already a given. If you don’t focus the discussion solely on how to make it work as well as possible, you’ll end up creating a negative wave of “maybe we should just stop.” Over the past year, we’ve taken a careful, thorough approach—making sure to check every PoC and all the important details—in order to aim for AI-Native. Once that decision is made, you can really sense the determination to see things through to the end.

Enjoyment accelerating our shift toward a truly collaborative environment with AI

@gomichan: Finally, what do you think will be crucial for Mercari as we aim to drive even greater transformation going forward?

@mark: I think it’s about looking at everything from an AI-centric perspective. You’ll hit a wall if you think of AI only as a substitute for something else. Instead, when you build processes with AI at the core, it’s crucial to start from scratch with the understanding that everything—your organization, your systems, and even people’s roles—will fundamentally change.

At that point, you really have to make the most of your in-house assets—your expertise—if you want to reach the next stage. I think Mercari’s real advantage is that the whole company is getting involved and moving things forward together. For example, having expertise only in legal or finance won’t easily get us to the next stage. But by effectively combining that with engineering expertise, it becomes possible to create systems that fundamentally change the way work is done.

@kimuras: I think enjoying what you’re doing is pretty important if you want to keep moving forward. Just like with studying, things get more fun as you start to understand them. Getting through tough times and finally reaching the stage where everyone across the company could actually use and enjoy AI as a tool—that was a real turning point. Going forward, it’s not just about enforcing things from the top down—it’s important for everyone to experience the benefits of generative AI for themselves and have fun imagining new ways to work.

@mark: I think it’s fantastic to see so many stories of people genuinely enjoying AI popping up everywhere. The more fun someone is having, the more they want to share it and spread the word—doing their best to get others hooked too. (laughs) It’s not just the engineers—teams who don’t have hands-on engineering experience like HR, customer service, legal, design, and more are now bubbling with excitement and saying things like, “When I tried this myself, I realized I could do all sorts of things!” That sense of fun and energy is really noticeable.

@gomichan: At the end of the day, the fastest way is to just have fun, dive in, and get genuinely interested, isn’t it? Hearing what you have to say, it really struck me that the reason we’ve come this far is because people at Mercari got genuinely interested in this technology and had a blast exploring it.

Pictures: Tomohiro Takeshita