Here at Mercari, most members have already started using AI in their daily work as we move toward becoming AI-Native. However, when we talk about “AI-Native,” there may be some who do not fully understand what Mercari has its sights on for the company or its members.

So, how are Mercari’s AI-Native members utilizing AI? To uncover the answer to this question, Mercan has launched this interview series titled “Yurino’s AI Question Corner,” featuring Yurino Horiuchi (@Yurino)—a Mercari intern who currently works on the AI Task Force and will be joining us as a permanent member soon!

For this fourth volume of the series, @Yurino spoke with Hitomi Suwa (@suwachan), who is responsible for data analysis and the promotion of AI usage within HR. After transferring from her previous role as a product data analyst to work in HR, she became involved in a Mercari initiative to create a fairer working environment through the examination of data. The project sought to visualize and correct the gender pay gap.

For analysts, is AI about democratizing data analysis or enhancing it? How should we use the time we save? Imagine having to redesign your work as the manager of ten gifted University of Tokyo graduates. Where would you start? We look at these and other topics in our interview with @suwachan, who supports Mercari using both data and AI!

Whether you have the will to improve using data affects outcomes

@Yurino: So, what was it that prompted you to take on the challenge of AI usage, @suwachan?

@suwachan: For me, the trigger was the emergence of ChatGPT 3.5 around 2023. I felt compelled to ride the wave, so I took an AI management course organized by the Matsuo Lab (also known as the Matsuo-Iwasawa Lab) through the University of Tokyo. I started learning about various AI utilization cases in different industries by listening to lectures by guest speakers about how they were using AI.

Also, when I heard that we were establishing the AI Task Force at Mercari last year, my hand shot up immediately. Now, I am working on two fronts: I have my work as a data analyst, and I promote AI in the HR field as a member of the AI Task Force.

@Yurino: In what situations do you usually use AI?

@suwachan: For the most part, I use it to write code for data analysis. I use AI in almost every aspect of my work—for example, generating code and getting code suggestions on Google Colaboratory (Colab), checking the logic of analysis reports, and learning new analysis methods when I’m unsure about something. Thanks to AI, I feel that I can now do things that were previously limited by my time and skill barriers.

@Yurino: AI has become an indispensable tool for data analysis, right? So from your perspective as a data analyst, would you say that the emergence of generative AI is more about democratizing analysis—making data analysis more accessible to non-experts—or about enhancing analysis by helping those already working in data analysis to take their work to the next level?

@suwachan: I would say both, but personally, I feel that it’s doing more in terms of analysis enhancement. Thanks to AI, I can immediately ask and get the answers to questions about things I don’t understand. I’m able to grasp ideas and acquire knowledge significantly faster than before. This is why I believe AI helps enhance my work by allowing me to try new methods and by increasing analysis speed so I can take even deeper dives. If you feel driven to do this kind of thing, AI can be a great help, but on the other hand, if you don’t feel any drive to uncover things by using data—even if you have AI at your disposal—the technology will just be a wasted resource. The democratization of technology does not march ahead simply because AI exists; it requires the improvement of data literacy and the fostering of a data-driven culture.

Of course, there are also constraints related to security and rules, but I feel that without an organization’s appetite to make data-driven improvements to our work, democratization won’t spread. That’s why it’s important for our organization to define what data we have, how it’s organized, and how we’ll use it. I believe that feeling driven to make further improvements using data is important.

Presenting invisible unfairness in a way that everyone can agree on

@Yurino: I’d like to hear about your career as well. Why did you choose to become a data analyst in the first place?

@suwachan: As a new graduate, I worked in search engine optimization (SEO) and web marketing. My experience in these fields taught me the importance of data and analysis, steering me toward becoming a data analyst. I have a liberal arts background, so at first, I wasn’t in love with the numbers or analysis work, but as I continued, they grew on me. Moreover, because I learned analysis from scratch and from the perspective of a liberal arts major, the path I took to understand things gave me the tools to explain what I learned to other members, which made my experience even more enjoyable.

After that, I wanted to dive even deeper into data analysis, so I changed jobs to become a product data analyst.

@Yurino: I understand that following your time as a data analyst, you moved on to working in HR. Why did you decide to take on the HR field?

@suwachan: After gaining roughly two years of experience as a data analyst in the product area, I felt a certain level of achievement and started to again ask myself what I wanted to do next. At that point, I realized I wanted to solve issues related to workstyles and people’s careers, so I made up my mind to become an HR data analyst and applied for a transfer.

@Yurino: So you kept using your skills as a data analyst but took on the challenge of a new field. What kind of work are you doing in HR?

@suwachan: I use data to shed light on the current working conditions at the company and help create better HR policies and workstyles. One of the best examples of my work was a project where we used data to visualize and help close the gender pay gap among our employees. We strive to present even subtle inequities that might otherwise go unnoticed, in a way that everyone can recognize.

@Yurino: It sounds like the numbers and data serve as evidence and create a foundation for fairness. As AI becomes more integral to HR going forward, how do you think analysts can ensure fairness is maintained?

@suwachan: Even if AI is integrated into the analysis process, I believe the ultimate responsibility lies with human analysts. That is why I value professional ethics above all else. Humans are inherently biased creatures. I recognize the existence of biases like automation bias (overly trusting AI suggestions) and survivorship bias (overvaluing success stories and neglecting proper risk assessment). With these things in mind, I strive to create fair evaluations and systems while remaining aware that I am also susceptible to these biases. Additionally, all AI output must always be carefully reviewed by humans. Since reports generated by AI may contain biases, it’s crucial to review the content.

Don’t just stop at achieving efficiency: Converting free time into value

@Yurino: I understand you host the “HR AI Office Hour,” a meeting time where you offer HR members consultation sessions on using AI within the company, right? Could you share what prompted you to start this initiative?

@suwachan: I believe anyone can relate to the situation where they see some kind of work that could benefit from using AI, but don’t know where to start. Alternatively, they try something and run into a wall along the way. We launched the office hours initiative in July of last year, believing that having a place where people can easily seek advice would help accelerate the adoption of AI. I wanted to create a place for people who are new to AI or who hit a wall to ask questions so that they don’t just give up because they don’t understand something.

@Yurino: Do the members who come to consult you share any common issues?

@suwachan: The most common issue is that they don’t know the extent to which they can input data into AI. It’s only natural to feel apprehensive about this, since HR handles highly sensitive private information. That’s why we work together to confirm what’s acceptable from a security standpoint as we explore ways to use AI. And even if I get stumped by a technical question, I use AI to find solutions for our members.

@Yurino: What security considerations do you keep in mind when actually using AI?

@suwachan: One is to protect security and privacy. Mercari has security and privacy teams, as well as an AI governance team that has established guidelines for AI use, so following these guidelines is essential. Another is to always review the AI’s output yourself. I make sure I can explain the results myself and take responsibility for them. Since simply saying “AI wrote this” could undermine people’s trust, I make a point never to skip these two steps.

@Yurino: What are the things you have found to be necessary to the process of promoting AI?

@suwachan: I believe it’s important to think about how to use the time we save by using AI. While increasing response speed and being able to attend meetings are important matters, I think AI truly adds value only when the time saved is used to focus on higher-value work.

@Yurino: You’re right that it’s important to consider how to use the time gained through efficiency.

Using AI to engage with people

@Yurino: What areas of HR work do you think are compatible with AI? Are there areas where it is harder to apply AI?

@suwachan: Currently, the areas compatible with AI tend to have the following traits:

1. Areas with a large amount of data

2. Areas dealing with unstructured data

3. Areas where the accuracy requirements are not too high

Recruitment involves a large amount of text data, such as job postings, resumes, and interview notes. This means there are many opportunities to leverage AI for screening, matching, and scheduling. Although a person makes the final decision, this is an area that we can investigate to see whether AI can be used and how to improve its accuracy.

On the other hand, it’s harder to use AI in areas where accuracy is critical. For payroll tasks such as attendance and salary calculations, where mistakes are not acceptable, more reliable automation technologies like robotic process automation (RPA) may sometimes be more appropriate than AI. I think the jobs of roles like human resources business partners (HRBPs), who use dialogue to enhance the quality of HR processes, are still a challenge for AI to handle on its own.

@Yurino: How do you think AI can support an AI-Native HR organization in terms of data infrastructure?

@suwachan: For that to happen, I believe it’s important to organize HR data so that AI can leverage it safely. To achieve this, we’ve been organizing and structuring data into AI-readable formats, while advancing designs that ensure the right data reaches the right people, with governance as a foundational principle. Going forward, we aim to further strengthen this foundation.

@Yurino: What challenges would you like to take on in the future?

@suwachan: I want to take things a step further by redesigning and rebuilding our work to be even more AI-Native. I read a thought-provoking question in a book about designing work with AI in mind: “Suppose ten new graduates of an elite school like the University of Tokyo joined your team. What would you have them do?” I think this question perfectly captures the essence of the issue at hand. An AI agent is a near equivalent to having a team made up of gifted University of Tokyo graduates work for you 24/7. We haven’t redesigned our work with that premise in mind yet, so that’s the next challenge I’d like to take on. It’s easier said than done, but unless we take on these challenges, we won’t be able to become a truly AI-Native company.

@Yurino: As we redesign our work, it seems the role of HR itself will also change. What do you think will be different about HR when AI becomes more ubiquitous than it is now?

@suwachan: As long as there are humans, I believe that jobs that involve dealing with people will remain. In addition, I think that as we work to coexist with AI, various issues will arise, so roles in which people have knowledge of both AI and HR will be important for solving those issues.

@Yurino: What does being an AI-Native person mean to you personally?

@suwachan: To me it means someone who believes in the potential of AI and who can fully harness its capabilities. The belief that AI-Native people within the company share is the idea that there is virtually nothing AI can’t do. I believe the AI-Native mindset is especially strong with individuals who can embrace challenges without limitations and identify opportunities to leverage AI in virtually any field.

@Yurino: Do you have anything you would like to share with members at the company who are gearing up to take on further AI challenges?

@suwachan: I believe AI usage is an area that cannot forge ahead unless it is driving change. It’s crucial that we stay fired up about the here and now, and not waste any opportunities. Let’s build the momentum together!

@Yurino: Thank you very much!

Check out other articles from this series!

Photography: Tomohiro Takeshita