Back in 2016 when he was assigned to the Distribution Solutions Division as an applications engineer, Fujita initially worked on systems development for one of Japan’s largest retailers. For the first two years, he spent his days getting thoroughly versed in the basics of systems engineer work, such as programming in Java and batch processing construction.
“I’d majored in college in statistical analysis and mathematical optimization, and so I had the desire to work as a data scientist one day. However, as an engineer I first got to experience how large-scale systems work and how they’re created, and this really shaped me into who I am today. That’s because without knowing the mechanisms by which systems work, you can’t integrate analysis results into practical operations.”

A turning point came during the spring of his third year there. An “AI-driven purchase order” project was being launched, which involved using AI to automate and support a business’s ordering process. At the time, the project was still in the PoC (proof of concept) stage. Fujita volunteered to be a core member of the project, which had begun with just two persons, and rather than simply conducting interviews about the ordering process at the store, he put on an employee uniform and experienced firsthand the work involved in placing orders. As a data scientist, he took on every manner of role, such as constructing analytical models, doing systems implementation, and even assisting with the on-site rollout.

“I would create the system and handle data analysis myself. I’d go to their stores nationwide and get direct feedback from the people there, and then I’d use that feedback to make further improvements. It was literally a setting where I had to do everything myself, but this experience significantly changed my life. That’s because I was able to feel in a tangible way both the struggles and the joys of making those analysis results a formal part of on-site operations. Actually, when I applied to join NRI, I even wrote that I wanted to use data analytics to help make it more convenient for customers to get what they need quickly, and to cut down on food waste. Being able to bring that idea to fruition as a concrete solution in the form of this AI purchase order project became a major source of confidence for me.”

Fujita first became familiar with the difficulties and fascinations of how to adapt technologies to business practices while doing research in his college days. At the time, Fujita was conducting joint research to create employee shift schedules at a large-scale hotel’s restaurant by predicting demand and using mathematical optimization.
“This wasn’t a case where simply getting enough people on staff would suffice. The issue was how to go about increasing computation speed while taking into account a complex combination of factors, including the skills, compatibility, and attendance of the individual staff members. We were also keen on avoiding any drops in service quality. For instance, we focused on how to factor into our algorithm things like ensuring that repeat customers get served by the same staff member, accounting for any specialties at the restaurant that were hard to translate into data, or other aspects of hospitality. That experience of conducting trial and error together with my counterpart at the enterprise has served as a foundation for my current work.”
This perspective of bridging the gap between technological capabilities and on-site practical operational processes was the starting point for Fujita’s present efforts.

“Generative AI is helping individuals achieve greater productivity to a certain degree. However, there still remains a high barrier to the domain that lies just ahead —in other words, the phase that involves incorporating AI into actual operational flows and comprehensively utilizing multiple AI agents as an organization. The issue here is how to link and integrate the company’s proprietary knowledge and data, which general models haven’t learned, with existing systems, and then run operations with governance and security being ensured. A lot of companies are feeling that gap right now.”

Even trickier than applying that knowledge is connecting it with practical operations. “Making it fit with your company’s business practices and organizational climate or culture is much harder work than entering industry-specific data is. It involves the front-line divisions with expertise in business operations, as well as the IT divisions that manage systems. Having one or the other set out determined to do it themselves is not sufficient to bridge this gap. Rather than simply applying AI to current operations as-is, we need to consider together with clients what ‘AI-driven operations’ would mean, premised on the existence and use of AI. It’s a rapidly changing field, and that makes it essential to engage in multifaceted efforts, by agilely verifying what you’re able to accomplish right now, and at the same time making forward-thinking plans for the future.”

Fujita defines his own role not as a specialist in any given technology, but rather as “someone who’s always searching for optimal ways of solving problems.”
“IT, data analysis, generative AI, and AI agents are all tools for achieving a goal, and so I think they’re merely the ‘how’ part. You always have to keep asking what the ideal solution would be in a given circumstance, and then work to create something that has never existed before while trying out new technologies with your own hands. Enjoying the process of working and thinking together with clients to address their needs will lead, in the end, to solving society’s challenges. I believe that repeating this process will help us realize NRI’s corporate philosophy to ‘prosper together with our clients’.”

Fujita says he even wants his own expertise to change and evolve, in keeping with the times and with new challenges. He looks beyond technological advances, with an eye toward achieving a more prosperous society.