Mercari’s data-informed AI transformation with DX

Mercari, Japan’s largest marketplace platform with millions of users, has never shied away from taking bold bets. When the company set out to become an AI-native organization, it was a natural extension of that culture. “Our leadership recognized early on that we were facing a major industry transformation with AI,” says Michael Galloway, VP of Platform. “To successfully navigate it, we needed visibility into how AI tools were being adopted and the impact they were having across our engineering teams."

Initially, to drive AI adoption, Mercari took an experimental approach, encouraging developers to use AI tooling they felt most comfortable with. “At the start, we weren’t trying to standardize anything,” describes Snehal Shinde, Director of Engineering. “We just wanted people to try different AI tools and see what actually worked for them. What helps a backend engineer doesn’t always help someone on iOS or Android, for example, so it was all about exploration in using AI tools to simplify the work.”

At the time, Mercari didn’t have reliable data about its developer experience—they didn’t have a way to see how engineers were working, what friction they faced, or how AI was changing their productivity. That changed with DX. “I feel DX provided us with a tool to actually look at the full developer experience and the impact of AI,” says Shinde. “It gave us a foundation to measure things like PR reviews and DORA metrics, and then later we started building our own dashboards to dig deeper into specific problems. DX was the core piece we needed to make data-driven decisions around AI tooling.”

Mercari’s teams now rely on DX data to guide every phase of their AI rollout, from experimentation to enablement to impact measurement.

1. Continuous experimentation and vendor evaluation. Early dashboards built in DX showed which AI tools were being used by which roles: engineers, managers, and even product managers. Over time, Mercari saw usage shift from GitHub Copilot to newer tools like Cursor, Devin, and beyond, reflecting how quickly the AI landscape was changing. “Different segments of developers used different tools,” says Kengo Fujishiro, Senior Engineering Manager. “DX helped us visualize those shifts and see which tools were most effective for each group. That insight allowed us to support each community differently.”

2. Increasing adoption through targeted enablement. The dashboards also exposed pockets of low adoption, enabling targeted intervention. “We had a couple of teams saying, ‘We have many AI tools, but we don’t know how to actually use them,’” says Shinde. “Because we could observe each team’s metrics in DX, it was easy to flag where adoption was lagging and start asking why.”

Those conversations uncovered a range of challenges. Some developers weren’t sure how AI fit into their daily workflows, while others struggled with complex codebases that made integration difficult. To address this, Mercari organized a series of lightning talks where engineers shared how they were using AI in practice, helping others find real, actionable use cases. For teams facing architectural barriers, those same discussions sparked ideas on how to evolve systems to be more AI-friendly. “We gradually saw that all team members started using AI, and each of them came out with their own creative way of using it,” Shinde remarks.

3. Understanding downstream impact. As more developers adopted AI tools, Mercari began to examine second-order productivity measures. At one point, Galloway pulled dashboard data from DX for an all-hands presentation about reliability and cost management. “I wanted to describe what the coming tidal wave is starting to look like in terms of the amount of changes in our codebase being drafted by agents,” he explains.

This visibility raised important questions: How do existing systems handle dramatically higher change volumes? If error detection catches 99.9% of failures with 10,000 changes, what happens with 100,000 agent-generated changes? “Being able to visualize those patterns and relate them to the other systems and metrics that we’re tracking in DX is critical,” says Galloway. “It helps us understand and communicate what adjustments we need to make as AI adoption increases. That’s especially important for Platform teams like ours.”

Now having achieved broader AI adoption, Mercari’s focus is shifting toward refining and sustaining impact. “Phase 1 was about 100% adoption,” says Shinde. “Phase 2 is about fine-tuning and making it real.” Galloway agrees: “The next frontier is understanding what AI agents mean for our systems and practices. With DX, we have the visibility to navigate that evolution thoughtfully.”

By combining Mercari’s strategy with DX’s measurement and insight, the company continues to advance its mission to integrate AI into every part of the software development lifecycle. “At the end of the day, what we want to do is ship impactful features to customers more effectively and with less friction,” concludes Shinde.