The Programming Health Nurses at Mazda: Inside AI Dojo and New MAX Organization

Shizuka Aramata, senior occupational health nurse. She has spent her career supporting employees in manufacturing and development through health consultations and guidance. Her motto: “I want Mazda employees to be known for thriving both during and after their careers.”

As the company’s health promotion department, we have health checkup data, questionnaires, and stress check results for all 24,000 employees. Some people get checkups twice a year, so we accumulate over 30,000 data points annually. We had plenty of data, but Excel was our only analysis tool. We could identify broad trends like “obesity rates have increased in recent years,” but couldn’t go deeper. I constantly questioned myself: “Are we identifying the right issues?” “Are our initiatives actually effective?”

Then, the head of company-wide digitalization reached out. “You have lots of data at the Health Promotion Center, right? How about using AI to analyze it and improve employee health and well-being?” I was surprised. How did he know about our struggles?

When I learned that AI could predict health risks, I immediately said, “Yes, please!” Of course, I thought the IT department would do all the work, so it sounded perfect. But later we learned that we’d be doing the work ourselves. That’s when things really started to get intense!

We enrolled in AI Dojo in June 2024. AI Dojo is Mazda’s internal program where IT department employees serve as instructors, holding two-hour sessions each week for study and practical consultation. We started by setting up our work environment to run Python, then learned coding, database creation, and statistical concepts like distribution and normal distribution from our instructors. But most of it went completely over our heads.

The first lesson kicked-off with “principal component analysis.” When the instructors saw how confused we were, they created a custom textbook just for us.

The custom textbook AI Dojo instructors created for the health nurse team. The notes and sticky tabs show their dedication.

The employee health data we wanted to analyze is sensitive personal information. Even though the instructors are Mazda employees, we couldn’t show them the raw data. We had to anonymize it so individuals couldn’t be identified. They taught us the process, but since we couldn’t show them the data, we had to do everything ourselves. I thought, “This is not going to be easy!”

I’d never even heard of Python before, but things escalated quickly, and it was too late to back out. I threw myself into studying using Aidemy Program*, YouTube, and textbooks.

*Aidemy is an online AI and digital technology learning platform for companies provided by Aidemy, Inc. Mazda introduced it in 2022 to develop digital talent across all administrative and engineering departments.

Creating a database for all employees meant linking data that had existed separately to individual people. When we tried to consolidate ten years of data, checkup items varied by year, and testing methods and reference values had changed. Everything was inconsistent. We struggled to align it all, and when we finally thought we’d succeeded, there were only 500 records. Wait, where are the other 23,500??? We kept reviewing the code, checking this problem or that. After three months of trial and error, we finally completed it. And once the database was ready, the analysis was...

Instant! The instructors kept telling us the database is everything. Once you have that, analysis results appear in seconds.

When we received the AI Dojo invitation, one condition was to include young professionals in the team. Many people in the younger generation have strong IT skills, and they wanted us to tap into that potential. But younger staff alone might lack experience and misjudge the direction we need to go. So they wanted a mix of experience and youth. At the time, Tsuneshige had only been with us for six months, Nagato was mid-career, and I was the veteran. All three of us were Python beginners, but Tsuneshige picked up coding incredibly fast.

When I showed the analysis results to my senior colleagues saying “Look what the data shows!” they’d get excited and say “I knew it!” Shortly after joining Mazda, people from not just my own team but other teams started asking “Could you pull this kind of data?” It felt good to help everyone. That’s what kept me motivated to learn Python.

We used to analyze data individually, but now that we have a shared database, it’s saved us so much time. When data confirmed what we’d sensed from experience, it sparked everyone’s desire to do more. That chain reaction kept us motivated. We had so much to do and it was hard, but it was fun, right?

So much fun!

Mazda’s purpose is “enriching life-in-motion for those we serve.” To enable that, our goal as occupational health nurses is to use AI to identify what factors influence work engagement and presenteeism, essentially workplace productivity. People don’t always recognize how important health is while they’re working, but disease risk increases with age.

People are a company’s greatest asset. That’s why being able to pinpoint factors for health promotion based on real data from Mazda employees, not just general trends, is significant progress.

Actually, in April, 2025, we presented our findings on employee health conditions and health management analysis to Mazda executives. Our analysis showed that work engagement increases when people feel their job suits them, and when they have strong support from supervisors and colleagues. When we explained this came from real data on Mazda employees, we had the executives’ full attention. To turn these findings into concrete initiatives and improve our health promotion efforts, we now want to create department-specific analysis reports that generate automatically for each department head.

Beyond car manufacturing, Mazda has many different job roles and people working in diverse ways. As an occupational health nurse, supporting employee health at a company like this gives me real purpose. Our mission at the Health Promotion Center is to help everyone stay healthy and thrive in their communities throughout their lives, even after leaving Mazda.

Since joining Mazda, Masahiro Yoshioka has consistently worked on infrastructure and new technologies in IT. He works daily toward a vision where all Mazda employees use AI effectively and focus on work unique to humans. As AI evolves, Yoshioka explores fundamental questions: What can only humans do? What makes human contribution valuable? Together with Yoshioka and these questions, the new MAX organization has begun its work.

Mazda’s 2030 management policy calls for value creation through the collaboration of people and IT. We plan to improve business process efficiency by connecting AI with data. To make this happen, we’ve used Aidemy Program for internal IT training since 2022. But realistically, no amount of study helps without opportunities to practice. That’s why we launched AI Dojo.

My team already had members with AI skills who’d worked on AI implementation projects with production departments and factories, so they became instructors. Most instructors joined through mid-career hiring. We have 46 instructors working remotely across Hiroshima, Tokyo, and Kansai. AI Dojo includes both projects we lead directly and projects where people work hands-on themselves, like the health nurse team led by Aramata. We decide the support approach based on each project’s needs and requirements.

When Aramata’s team joined, AI Dojo had just launched and participants were mainly from manufacturing, so having health nurses there was unusual. We’re amazed by how far they’ve come.

Aramata reflects on her year in AI Dojo. A health nurse and an IT engineer sitting side by side for an interview, discussing the same topic. A scene unique to this initiative.

In September 2025, we established MAX (Mazda AI Transformation), a project office to drive cross-functional business transformation using AI. MAX is actively introducing generative AI in the areas of administration and engineering, working toward company-wide data organization and consolidation. At the same time, we’re expanding AI applications across all areas so generative AI can address each department’s specific challenges, dramatically improving productivity, efficiency, and speed.

As the health nurse team experienced, acquiring AI skills and building a track record takes considerable time. You need to learn what approaches don’t work and which ones fit specific situations. It’s a constant cycle of trial and error. But that process, including the failures, is the real asset, and the value lies in accumulating that knowledge internally.

None of this happens overnight. It’s painstaking work that requires perseverance, but we believe that developing AI talent internally will drive Mazda’s sustainable growth. By automating manual tasks and Excel work, reducing workload across departments, and helping all employees maximize their human capabilities, we can achieve this.

As AI adoption has gradually spread from AI Dojo, we’ve seen fascinating examples of application from challenges I had no idea existed. They are exactly the kind of solution that comes from people on the ground.

We have so much information inside the company which is unavailable to external AI. Mazda’s information exists only at Mazda, so the real challenge is determining how to apply AI to create value from it.