Greetings from the Mercan editorial team!

In May 2026, Mercari launched a company-wide program called AI Agent Day, calling on around 1,000 non-engineer employees to participate. Over the course of four months—including preparation and integration phases—all eligible participants will join in one of six learning session batches, with some batches running concurrently.

Each batch of the program takes about one month to complete. During the two-week preparation period leading up to the main event, participating employees will get hands-on experience with AI tools. Then on the day of the event, they will use Notion AI and Claude Cowork to develop their own AI agent skills. The main event will be followed by another two-week period during which employees focus on integrating these skills into their daily work. More than just a one-day event, this program is specifically designed to drive lasting behavioral change.

In this series, “Uncovering AI Agent Day in 10 Questions,” we will uncover the true face of the program in six articles, covering everything from the overall concept to reflections on each batch.

Ahead of launching this series, we’re using this article to introduce an overview that looks at what kind of initiative AI Agent Day will be and to pose our first two interview questions to the three project leaders.

Note: This article was composed and written using AI tools, based on notes and interview transcriptions. #AIWritersDesk #UncoveringAIAgentDayIn10Questions

Overview: What is AI Agent Day?

AI Agent Day is a company-wide program initiated by Mercari as it aims to become one of the world’s leading AI-Native companies. The program targets approximately 1,000 non-engineer members, including executive officers and vice presidents, and tasks them with changing their work experience using AI agents.

A key feature of the program is that it is designed to be a one-month process for behavioral change, rather than a one-day event. In the two weeks prior to the main event, participants complete their learning while engaging with the AI tools to be used on the day of the event.

On the day of the event, participants will get acquainted with the concept of AI skills by using Notion AI in the morning. In the afternoon, they will use Claude Cowork, a secure work environment installed on their computers, to actively develop skills directly related to their work. After the program, participants will engage in peer learning groups composed of seven to ten members from different divisions. This will help to reinforce their learning.

Supporting the retention process for participants are “AI Champions.” Approximately 20 champions are assigned to each batch. Members selected for their being approachable for consultation rather than for their technical expertise will accompany each peer learning group, supporting them for the duration from before to after the program.

Question 1: Why is AI Agent Day targeted at all employees regardless of their job type?

—This program is aimed at all members, not just engineers. Can you explain why it was made an inclusive initiative without limiting it to certain job types?

@ISSA: From the beginning, this program was designed with the premise of all employees taking part.

I believed it was only natural to distribute tools that all members, regardless of their job type or division, could use to improve results and productivity. Last year, chat-based AI tools like ChatGPT and Gemini were distributed throughout the company. Since then, AI has evolved even further, moving beyond a mere chat-based tool for bouncing ideas, to an approach where AI is able to autonomously drive toward objectives as an agent. This naturally led to the conclusion that anyone should be able to engage with agentic AI.

@shuuk: Examining our progress on an issue basis, I believe we have to quash the sense of treating this technology as an external contractor, which can arise from taking a hands-off attitude. An inherent imbalance arises between engineers and those in other roles when everyone has this attitude. What is obvious to engineers who use Claude Code on a daily basis may not be so apparent to people working in other roles. This difference often results in increased communication overhead and forces non-engineers to rely on technically minded colleagues when building AI agents. I have long felt that unless we fundamentally change this structure, the adoption of AI agents will never scale on a company-wide level.

After experiencing AI Agent Day firsthand, I realized that learning to use AI is a lot like learning to ride a bicycle. When you first start riding, you fall a lot, but then you gradually get the hang of it. Once you can ride, it feels so natural that you can hardly remember what it was like before. Just as people who have ridden bicycles can talk together naturally about cycling, I believe it’s important for everyone to master AI so that we can all communicate on equal ground.

Currently, there are divisions that have engineers stationed as enablers. But in the future, I think things will scale rapidly once we reach a point where we can handle certain tasks without relying on engineers. Now that we’re in an age where you can turn to AI to build something functional, I believe it’s important to reduce the information asymmetry among different roles and work together.

—I see. While full participation in AI Agent Day is a prerequisite, what was the intention behind making it a month-long program for each batch—including preparation and follow-up phases—instead of just a one-day hackathon?

@sae: A typical one-day hackathon often ends with people saying that they had fun and nothing more comes of it. Even when participants are recruited from across the company or from different divisions, it’s common for only the highly motivated to really gain something. Some of those who join just because they were invited often just try out the technology and then don’t touch it again after the event.

At first, I was concerned that this initiative would be meaningless if it ended after just one day. But when I started thinking of it as a month-long process of behavior change—with two-week lead-in and follow-up periods—I suddenly felt confident that it could actually work. Ideally, participants do 90% of the work ahead of time, so on the day itself, participants can focus on discussions based on what they’ve prepared. Through these ongoing discussions, they naturally gain insights, and before they realize it, those insights have already become part of their routine.

In addition, as we use AI in our daily work, I believe it’s also important to have someone to consult with. We’ve introduced roles like “champions”—people you can easily bring your questions to—and created small peer learning groups of seven to ten members, setting up systems designed to spark the same kind of casual, natural communication you’d get at a shuffle lunch.

Question 2: What exactly does “skill” mean in this context?

—What kinds of “skills” will participants create during AI Agent Day?

@sae: First, understand that a “skill” is essentially a concept similar to a reusable prompt. For example, abilities like writing blog posts, creating presentation materials, or preparing reports—if AI agents are equipped with these skills so that anyone can use them when needed, the skills become truly reusable. Moreover, by sharing the skills we develop with team members and the entire company, the potential for reuse expands from individuals to larger units like groups and divisions. I believe this approach is well-suited to our goal of becoming an AI-Native company or organization.

Another reason we adopted skill creation for AI Agent Day is its high level of reproducibility. We want participants to experience what it’s like to nurture their AI agent by continuously refining the skills they create.

@ISSA: There are stages in AI development: one is the co-pilot stage, where you bounce ideas in a chat environment, and the other is the agent stage, where the AI is given goals and it then acts autonomously to produce output. The concept of skills has become mainstream as a method for advancing to the stage of collaborating with agents. We want everyone to move from the co-pilot stage to the next stage—that’s why mastering skills is so important.

@shuuk: At this point, I think that many members are familiar with chat-based AI, but there are still many who aren’t yet used to AI agents that operate autonomously.

Even if they aren’t perfectly organized, by putting human skills and workflows into words and teaching them to the AI, we can refine them through trial and error.

By making these skills reusable, more people will be able to benefit from them. I hope everyone gets to have that moving experience.

References:

https://www.notion.com/help/skills-for-notion-agent

https://platform.claude.com/docs/en/agents-and-tools/agent-skills

—Listening to this, I felt very excited as a participant. In the next article of this series, we hope you’ll share more about the design process behind AI Agent Day!

Summary: Next time, explore the inner workings of the design process

In this article, we unfurled an overview of AI Agent Day and explored two key questions—why everyone participates, and what exactly the skills being created are—through the perspectives of the three planners.

In the next article (#1 in the series), we will take a deep dive behind the scenes, exploring everything from the conception to the execution of the project, together with the three planners who were directly involved. We’ll also share a full account of the February 2026 AI Task Force retreat held in Hakone, detailing the extraordinary speed at which consensus was reached with the leadership just one month later. The article will also feature the photos taken during our coverage of the event.

Greetings from the Mercan Editing Team. Over two days in February, approximately 60 members of the AI Task Force* (“AI-TF”) gathered for a retreat. *A cross-functional team driving business transformation across multiple domains through the use of AI.