The Customer Service & Trust and Safety (CS/TnS) organization at Mercari is in charge of ensuring safe and secure experiences for our users. As Mercari shifts toward company-wide AI adoption, CS/TnS hasn’t simply added AI to their workflow, but fundamentally reconstructed how they work with AI at its core. 

In this article, we talk to CS/TnS project managers @mika.u and @aguni as they reflect on how they undertook this initiative as an organization, where they encountered challenges, and what they prioritized along the way.

We hope this conversation will be helpful for others looking to adopt AI too!

Two dedicated members to ensure success

—To start, could you tell us what sort of framework was set up in order to work toward creating an AI-Native organization?

@aguni: AI adoption throughout an entire organization can’t be achieved by individual effort alone. From when the AI Task Force was first established, we were assigned as dedicated members to work directly with each area within CS and TnS to ensure that the entire organization got involved in the process.

Because we didn’t have other concurrent roles and responsibilities and could focus solely on promoting AI adoption, we had the capacity to deeply understand and organize workflows, and we could work as a pair to brainstorm and refine ideas for accelerating decision-making with AI.

As a result, it was much easier to rapidly iterate through the cycle of ideation, implementation, operation, and improvement.

Cataloging work processes to identify opportunities for AI 

—So then, how did you go about creating an AI-Native organization?

@mika.u: Even before selecting the tools, our starting point for creating an AI-Native organization was creating a “task blueprint.”

We cataloged and listed up all of the tasks performed throughout CS/TnS to visualize where AI could be used to substitute or support existing work. Doing this step thoroughly allowed us to make more clear and confident decisions down the line.

During this phase, we considered things like, “What tasks must be handled by a human?” “What tasks does AI excel at, like creating reports?” “What repetitive tasks do not require human judgment?” “What potential risks and quality concerns may there be?”

The key was not to start from the idea, “How can we implement AI?” but to work backward from the idea, “How can we change our workflows to create more value for our users and the CS organization?”

@aguni: After listing up all of the organization’s tasks, we then identified the areas of highest priority.

CS/TnS handles a broad range of work, so attacking all angles at once would disperse our efforts and weaken the results. Therefore, we prioritized the areas of highest impact and that would show the most visible results, focusing first on creating a “winning strategy.”

Demonstrating success from the start later became the fuel for the widespread adoption of AI.

The winning strategy for widespread adoption: space, praise, and reuse

—I see! So achieving initial success really helped pave the way.

@mika.u: We can’t truly become AI-Native simply by introducing AI tools and having those with the expertise maintain and manage them. In CS/TnS, we focused on creating a system that would encourage people to use AI, learn, and reuse what others have created.

1. Creating space for regular info-sharing sessions
   
   Introducing use cases at company-wide All Hands meetings, holding a CS/TnS AI Task Force All Hands meeting every two months, having leadership clearly articulate the organization’s commitment to becoming AI-Native, introducing ways people are leveraging AI that could be reused across the organization
   
   
2. Encouraging Slack posts and action through praise
   
   Encouraging members to post use cases on Slack, praising meaningful examples with emoji reactions and Mertips (Mercari’s tipping system), creating visibility around each team’s AI use cases, making it easier for those doing similar tasks in other domains to implement AI
   
   
3. Documenting examples of AI usage in a Notion database for searchability
   
   Documenting use cases in a Notion database to prevent them from getting lost, listing the “issue,” “result,” and “tools used” for each use case, making it easier for others to find solutions to similar issues they may be facing, and facilitating reuse

These three strategies together turn AI from something a few people are working hard to implement into a part of everyday work.

Transitioning from “trying” to “relying” 

@aguni: Looking back, the initial phase was really about getting people to actively try AI.

But after just six months, we could really see the difference throughout the organization.

When people see those around them using AI, they’re inclined to try using it too, creating a chain reaction. Tasks that used to take talking to another person, like brainstorming, summarization, and drafting, can now be discussed with AI, making things faster. With increased efficiency, people have time to tackle issues that were previously left unaddressed.

On the other hand, as AI becomes capable of more things, it can also feel as though our work is increasing along with the possibilities.

That’s why it was also important to make sure we stayed focused on not just increasing efficiency but providing greater value through AI-Native workflows.

@mika.u: Our goal was to achieve a state in which people don’t need to be told to use AI; rather they feel like they can’t work without it. Just as smartphones replaced feature phones, we want AI to become the standard and something everyone uses effortlessly in CS/TnS. I really feel like we’re standing at the doorway to that future.

The essence of AI utilization

—It sounds like you made great progress! Did you see or feel any examples of tangible results?

@aguni: Our initiatives in CS/TnS later won an MVP award within the company. What was significant about the award was that the recognition was directed not at individual efforts, but at the change within the organization as a whole.

There were several important elements that contributed to the initiative’s success: quickly defining the vision and articulating the strategy for AI agentization, focusing on how AI can transform the value we deliver to our users, rapidly iterating ideation, operation, and improvement to generate quantitative results, and strongly aligning between VPs and project members to effectively bring everyone on board. It’s easy for AI to remain a tool that only specific people are knowledgeable about and only some people use.

What winning this award demonstrated was that the essence of AI utilization isn’t about the technology, but about the organization. Real results manifest when commitment from leadership and passion from the frontlines align to generate an effective system.

@mika.u: We’re still on the way, but it’s definitely becoming more natural day by day for us to consider how things should work with AI as a given. We won’t let this change end as a temporary one or focus merely on efficiency. We will remain focused on the type of organization we want to become.

And, moving forward, we will continue to develop the form of value that we as the CS/TnS organization can provide to Mercari’s users.