How can we leverage AI in our day-to-day work?
You might be wondering this yourself, as this is one of the most asked questions today.

Here at Mercari, we are no exception. Up until recently, we used generative AI technology very sparingly. Now, we have experience creating an internal generative AI tool (Ellie) and adding AI features to our services, and the way we use generative AI within the company has changed significantly. Over 6 months, usage of Ellie increased from 61% to 75%, which indicates that employees are steadily finding more ways to leverage generative AI internally.

So, how has generative AI changed Mercari?
Gomi Hayakawa (@gomichan) looks back on the changes that occurred at the company from July to December 2024.

Related Article

Mercari Group’s Bold AI Initiatives Today and Beyond to Compete With the World’s Tech Companies

Read more

Generative AI usage at Mercari and issues in July 2024

When I joined Mercari in July 2024, generative AI was only being used in limited ways at the company.
While overall usage figures were not that low, employees were only using AI for machine translation and coding tasks. To many, generative AI was just a handy tool that translated for them; in other words, generative AI wasn’t being used to its fullest potential.

The first issue I wanted to tackle was increasing awareness of generative AI and LLM technology internally. Not many people were aware that they could use generative AI for anything other than machine translation or coding, so hardly anyone was using it for day-to-day work. Another problem was that, because members didn’t know much about how AI worked, they had very high expectations for it—so high that they got frustrated when it didn’t produce the results they were hoping for or when it output false information.

Change by the numbers

I’m pleased to announce that things changed a lot over the last 6 months of the year!

First, more and more people are now using generative AI. Looking at the monthly figures for our internal generative AI tool Ellie, the usage rate jumped from 61% to 75% in the span of 4 months. Also, we can see that retention rate (the rate that members revisit Ellie) over 30 days is steady at more than 80%, so the majority of members have used Ellie more than once. The number of tokens being processed is also increasing (i.e., there’s a higher volume of input and output), which likely means that members are now using Ellie in combination with data for more specific tasks.

We also encourage members to use Gemini for Google Workspace alongside our in-house generative AI tools, so the actual utilization rate of generative AI is even higher.

Related Article

The Generative AI/LLM Team’s Goal of Creating a Platform Equipped with a Wide Selection of AI Tools: Exploring the Intent and Background Behind Mercari’s Internal AI Environment “Ellie”

Read more

Organizational issues that generative AI can solve

Based on the company’s progress with generative AI over these 6 months, I think the issues related to internal tasks that would benefit from using generative AI can be categorized as follows.
I also believe that our initiatives over this period have helped upgrade the potential issues in the bottom row to the apparent issues in the top row.

For the apparent organizational issue in the top left cell, the appropriate teams are steadily working to solve this.

For the apparent individual issues in the top right cell, given the highly individualized nature of each issue and the large number of members involved, an approach focused on helping individuals understand and utilize generative AI would likely be effective. Empowering members to help solve the issues in the top right cell is the job of my team, which was created in December 2024. Changing the way one person works might not have the biggest impact, but at a company of around 2,000 employees like Mercari, saving everyone 1 hour per month would mean improving work efficiency by 24,000 hours per year. With that extra time, we can focus more on our users and growing the company.

I expect that we’ll need to take the following steps to incorporate generative AI into our work.

This process will involve, first, leadership committing to leveraging generative AI, and then creating the right usage environments to remove any blockers, increase recognition, and accumulate more use cases. The environment surrounding generative AI is constantly changing, so we might have to tweak and repeat this process.

Initiatives

Our initiatives over these 6 months involved the following actions.

First, leadership showed strong commitment to generative AI. The strategy roadmap for the entire Group clearly states that Mercari will use AI products and incorporate AI into operations to enhance productivity. In order to achieve this roadmap, AI usage has been incorporated into division OKRs.

It has also been important for leadership to take the lead in utilizing AI as it becomes more integrated into our work. For example, leaders are actively leveraging AI by regularly setting agenda topics related to AI/LLMs in management meetings, using generative AI for summaries and follow-ups so that members can catch up on information related to management meetings more quickly, and making preparations to use generative AI for transcription, which was previously done by humans.

Next, we worked on establishing an environment in which we can use AI. For this step, I was conscious of not just preparing the right tools and information but also removing any blockers that would stop members from using AI more than once. For example, not knowing whether certain information can be input into the AI or not is a big blocker, so I wanted to create clear guidelines that help members with this. Of course, working with our dedicated internal generative AI/LLM team was crucial, but other teams were also enthusiastic about helping out. Teams that deal with risk, security, compliance, ethics, and legal, which could potentially hinder guideline updates, actively considered and experimented with how members would be using AI, so the guidelines turned out really well. The environment surrounding generative AI changes day by day, so we want to regularly update these guidelines so that they don’t hinder development.

We also actively implemented measures to widely promote the use of AI. The most effective action in this step was not something flashy, but the steady, grassroots efforts performed by members every day.

We actively shared information about AI and use cases during all-hands meetings and posted information in open Slack channels. For more casual interaction, I created a channel where I can post random thoughts and share information about generative AI that I find on social media. Around 350 people have joined my channel so far. I’ve also been holding a lot of internal study sessions called Lunch & Learns, which are an important part of Mercari’s culture. From September to December, I held 19 study sessions with a total of 1700 participants.

These initiatives are helping to shape an “AI as the new norm” culture, and members are starting to choose generative AI as a tool to help them work.

Our goal for 2025 is for individuals to leverage generative AI in teams, not only to improve work efficiency but also to provide value to the people who use our product. For my next Mercan article, I’d like to introduce specific use cases at the individual level.

Future outlook

There is still much to do to solve the individual issues I mentioned above, and we will continue to promote the use of generative AI at Mercari. With this in mind, we have set the team’s mission to “take full advantage of generative AI to maximize value proposition.” Our goal is to discover the potential of generative AI and share our insights with our users and everyone else related to Mercari to create a world where everyone can enjoy the benefits of generative AI.

When it comes to using generative AI in business, attention tends to focus on cost-cutting. However, I believe that the true potential of generative AI lies in utilizing the space beyond—that is, how we can use the extra time we gain.

Generative AI will fundamentally change our values. People got used to mobile phones changing to smartphones, and I think generative AI will be treated the same way. Among all this change, we will strive to ensure that Mercari remains a company that is flexible and continues to improve the user experience. To achieve this, all employees will work toward understanding and mastering generative AI for the future.

Author: Gomi Hayakawa