With the advent of generative AI, a wave of business transformation is happening across all industries, and the specialized world of corporate legal affairs is no exception. Over the past year, Mercari’s Legal Team has made rapid progress in its use of AI, establishing it as an indispensable tool for work.

How did they go from not knowing how to use AI at all to incorporating it into their daily work?

In this article, we sat down with Yuka Ochiai (@Yuka), Ayako Seya (@seya), and Teitei Ryu (@avaliu) of Mercari’s Legal Team to talk about their journey of overcoming psychological hurdles to AI and involving the entire team to promote its use, including concrete initiatives and strategies.

---

Starting off with uncertain exploration

—To get started, what was it like before the Legal Team started leveraging AI?

@seya: Our use of AI began in May 2024, when I participated in a workshop organized by the Corporate Engineering Team to identify areas where AI/LLMs could be applied. Honestly, that’s when I first learned that AI could be used for work.

After the workshop, I was asked to share what I had learned with my team, but I had a hard time processing the information and also presenting it in a clear way, so I was rather stuck. As of July 2024, a majority of the team was interested in AI but had never used it.

—So you started from scratch.

@Yuka: Yes. The workshop itself was very useful, but it was also very important for everyone to have a concrete idea of how they would use AI in their daily work as the next step. To get things started, in August 2024, I held an “AI seminar for beginners” for Legal Team members with the goal of lowering the psychological hurdles to using AI.

Eye-opening experiences breaking down psychological barriers

―What kind of seminar was it?

@Yuka: The concept was to “become friends with AI.” The goal was to demonstrate how AI can make work easier with only two words by introducing short and simple prompts that anyone can use immediately, like “summarize this,” “translate this,” and “proofread this.”

When it comes to new technologies, everyone is confused about what to do at first. That’s why I think it’s most important to set the bar as low as possible and get people to take the first step.

—How did the participants react?

@Yuka: I could see their faces light up as they discovered all the things that AI could do. We ran a survey after the event, and many people said they were impressed. I believe that this eye-opening experience was a major factor in changing everyone’s mindset toward AI. This seminar helped many members start using AI every day, and I felt that the psychological hurdles across the entire team lowered significantly.

Bringing ideas to life: AI/Automation Hackathon to boost enthusiasm 

—What was your next step after the change in the team’s mindset?

@Yuka: Once everyone had started to get the hang of using AI on an individual level, we decided to take on a new challenge: the Legal & Governance Division AI/Automation Hackathon. The goal of this hackathon was to go one step beyond individual use and, as a team, take on the challenge of improving productivity by further streamlining work and creating new value through AI and automation.

—Was there a particular reason you chose the hackathon format?

@Yuka: The key point was that after two weeks of development, we held a presentation event where each team showed off prototype demos of their ideas to a panel including Mercari VPs. I think that presenting in front of VPs made participants want to look as good as they could, so it naturally boosted their motivation more than it would have if they were just presenting to the people they usually work with. It was a good opportunity for teams to learn from each other, because everyone was thinking, “If I’m going to make a presentation, I need to create something that works.”

This initiative resulted in prototypes of 17 practical ideas.

Demos of the prototypes born out of the hackathon (Note that the checklist and terms of service used in the video are fictional)

Implementation into everyday work

—So, what happened to the ideas that came out of the hackathon?

@Yuka: In the next phase, we launched Project Everyday AI. This project aimed to fully implement things like the prototypes of the AI applications created at the hackathon into daily work. We hoped to embed AI in our daily work and create a state where we’re using AI every day.

@avaliu: I also participated in this project. We aimed to address concrete team needs, like wanting to organize emails with law offices by topic. In the process, @Yuka gave some advice: first, just start by creating an MVP (minimum viable product). From there, through trial and error, we were able to complete a mechanism. I learned the importance of first creating something that works, no matter how small.

---

Formulating an AI-Native roadmap to envision the future

—So after that, you created the AI-Native Roadmap. Could you give a summary of the roadmap and explain its background and purpose?

@Yuka: We believe that the advent of AI will significantly change the way people work. Routine tasks will be left to AI, and humans will shift to higher-value-added tasks such as strategy formulation, decision-making, and networking. Applying this mindset to our work, we created the AI-Native Future Vision, which shows what we think our work will look like in three years with the introduction of AI.

In the AI-Native Future Vision, we broke down our main work into small tasks, categorizing them into tasks that will be left to AI, tasks that humans will continue to do, and tasks that humans will be able to do in the additional time they will have once some tasks are handled by AI. For example, if the task is to create guidelines for the use of generative AI, AI will be in charge of collecting information and creating a draft of the guidelines, while humans will review the guidelines, make adjustments internally, and plan strategies in the time they have available. We actually had AI come up with ideas for the AI-Native Future Vision, using a prompt like the following: “Break down ____ work into small tasks and indicate the tasks that AI will perform and the tasks that humans will perform. For the tasks that AI will perform, suggest ways that we can use AI.” 

Basically, we did the same thing that the AI Task Force is currently doing across the company, but we did it first. In order for us to achieve this vision, the AI-Native Roadmap lays out in concrete terms what we will have implemented in six-month intervals.

—Why did you think the roadmap was necessary at this time?

@Yuka: The AI applications we created through the hackathon and Project Everyday AI only covered part of our work, because we started with the things that were easy to implement. We needed to think systematically about how we could fully automate our work using AI, so we created the AI-Native Roadmap with the goal of getting everyone on board.

Horizontal deployment to increase output tenfold or more

@Yuka: In addition, we have consistently focused on deploying our results across the organization. From the outset, the Legal Team prepared all the information needed to reproduce our AI applications and our expertise about promoting AI, with the intention of sharing it with other teams. We published all of the information for the seminar, hackathon, Project Everyday AI, and AI-Native Roadmap internally and encouraged everyone to “steal our knowledge!” Under the slogan of “let’s start a wave of AI from Legal to the whole company!”, we’ve always been thinking about achieving 10x and 20x results by deploying our results broadly beyond the Legal Team.

@avaliu: As part of those efforts, we set up an opportunity to share knowledge such as how we used AI to create e-learnings during an internal lunchtime study session. After the presentation, we received a message from someone in another team saying that they would love to use a method we presented in their team as well, showing that we were able to form collaboration between teams. I feel that storing this knowledge in a way that makes it available to the entire company and deploying it horizontally is very important.

Skills required of legal departments in the age of AI and future challenges

—After all these initiatives, how has your team’s workstyle changed?

@seya: The introduction of Hiyochan, an AI chatbot that handles inquiries within Mercari, was one of the major changes. When we first introduced it, we felt like we were fumbling around in the dark, but through repeated improvements, we were able to entrust AI with the first response to questions from employees regarding internal procedures. This allows us to focus on tasks that require more specialized judgment.

@Yuka: Our Generative AI Usage Guidelines, which I worked with other relevant teams to create, are very long. It’s a lot of work for employees to read the whole thing to find the answer to their question, and it’s even difficult for us to craft answers based on the guidelines. So we loaded all this into AI and created an AI-powered Q&A page (using NotebookLM) that we ask employees to try using first. This saves employees a lot of time by not having to read through a long document, and also saves us a lot of time since we can just link people to the Q&A page.

We also use AI in the field of patent applications. When considering formulating responses to office actions from the Patent Office, we have been able to use AI to greatly speed up work that had previously cost humans a lot of time.

@avaliu: In my work, I frequently use AI for research and other purposes on an individual level. Since there is a risk of hallucination (a phenomenon where AI generates plausible-sounding false information), we do ultimately check the text with our own eyes, but the research results that AI outputs are helpful for us to reference as we draw conclusions. Recently, we’ve also been using AI in our research on the Subcontracting Act; we uploaded all relevant laws, regulations, and guidelines to an AI, so now we can give it specific transaction conditions and ask whether the transaction is subject to restrictions or not.

In addition, we built a mechanism within the team in which simply using a specific reaction on a Slack message automatically summarizes the knowledge contained in the message and stores it in a separate spreadsheet. It’s a very convenient mechanism that automatically turns everyday interactions into assets.

@Yuka: The skills required in the AI era are definitely changing. For example, in corporate legal affairs, the ability to write text from scratch was important in the past, but now we will be expected to collaborate with AI and produce high-level output. I think that the ability to master AI will become essential.

―Can you tell us about any challenges that will need to be addressed going forward?

@Yuka: One challenge that I am particularly interested in is the training of new talent. Up until now, new talent in the legal field have been able to improve their skills by doing simple tasks and learning through repetition, but going forward, AI will take over those tasks, so we will need to think of new ways to train junior employees.

@avaliu: It’s also important to determine which AI will be used for which tasks. Each AI has its own strengths and weaknesses, so we’ll need to understand more about them and do some trial and error.

AI isn’t hard; it’s fun!

—Lastly, could you give a message to those who are looking to start leveraging AI in their work?

@Yuka: AI is a fun tool. And in order to change an organization, it’s essential for someone to be enthusiastic about that fun and influence those around them. First, share the experiences you found impressive with those around you. I’m sure that those experiences will be a powerful engine for changing your team’s mindset.

@avaliu: I think a lot of people are worried that AI will take away their jobs. However, with the spread of AI, new jobs and fields will inevitably be created. I think that rather than being afraid of AI, it’s very important to switch mindsets and find where you can best perform in a new area.

Photographer: Tomohiro Takeshita