Judgment Lineage

Project manager for core systems at financial institutions. Make what has been decided work, on time and without failure. Requirements came from above. There was almost no discretion to decide what to build.

Do not equate “working correctly” with “working as decided.” The latter comes first. In core financial systems, the certainty of execution that does not betray what was decided is itself the value, more than the quality of the decision. That discipline was learned here.

What was gained here was not a template for execution. The muscle to make a specification work and the muscle to question the specification itself are different. Without training the former first, the latter only spins in place.

AI worked exactly as specified. But was the specification right? Only someone who has learned implementation in their bones can see the gap between these two. Later work in decision-making governance design begins here.

The work began with development for membership and financial services: points, payments, securities, insurance, and FX. Member trust, financial risk, business, UX, and data — in domains where unlike things intersect, many lines must not be crossed.

The instruction from then-president Inoue, facing the accumulating member data, was: “Build an engine that knows our customers.” It led to a recommendation engine patented as No. 5095794 and a Hadoop platform of several hundred machines. That work led to responsibility for data solutions and company-wide data strategy. The question moved up a level: from making something work, to designing a system that continues to create value.

Do not perform individual analysis tasks one by one. Invest in the foundation that continuously processes large volumes of data. This decision was led directly. The question was not technology selection; it was a systems-design decision about who would continue to entrust what to that foundation.

A system’s value is not decided the moment it is built. It is decided over continued operation by the line between what to entrust to machines and what to retain for human judgment.

More important than the individual achievement of a patent is the habit formed here: design from the start for both systems that should scale and decisions that must not. How far to entrust to AI, and what to retain as human decision-making, is structurally the same question as today’s work.

Technical Director at the Data and Technology Center. Client interests, internal technical constraints, and the reality of the execution team. The work remained in the place where demands from different positions collide. The runner-up prize at the Tableau Viz Contest was recognized not as a technical demonstration, but as visualization that enabled management to decide.

Another responsibility was investment and alliances: investments and alliances with Japanese and international data and technology companies, including Treasure Data and Intimate Merger; drafting contracts, explaining them to the investment committee, and coordinating with finance and legal. The work identified structures the company lacked across business, data, and technology, then supplemented them by working with external parties. Management made the final decision; the role here was to accumulate the work and relationships until management could decide.

A technically correct proposal and a proposal management can actually decide on are different. The latter is the one to pursue. This decision was made repeatedly. The quality of visualization is only a means. The purpose was always to enable someone to decide.

Finding an answer and translating a question into a form in which the person who must decide can do so are different capabilities.

Later, after independence, this capability connected directly to recasting a stalled comparison of technology options as the one question management had to answer. AI can substitute for analysis and lists of options. Designing who should decide, and how, remains human work.

It began in July 2017 with a newly created Data Utilization Promotion Office reporting directly to the president. It was a team of one. A title did not come first. Listening across the company, organizing data, and identifying marketing issues — understanding the situation was the first job. Five months later, two specialists in data and digital marketing joined from outside.

The organization never moved from its position under the president. What changed were its name and its remit. In February 2018 it became the Marketing Division, taking the Advertising Department from PR and creating Digital Marketing and Data Platform departments. The aim was to strengthen the front: bring scattered customer touchpoints together in integrated marketing. In February 2019 it became the Digital Transformation Division, returning Advertising to PR and bringing in the IT Systems Department. The aim was to strengthen the base: connect with core systems and build internal digital literacy. First secure the front, then the foundation.

This was not holding several jobs at once. One organization that began with one person expanded, changing its name and remit with its results. What did not change was its position under the president and continued hiring of digital talent. Responsibility shifted from the success or failure of individual projects to the structure by which the organization makes decisions.

There was another crossing at this stage. Finance, internet services, and a comprehensive advertising agency — until then, all dealt in things without physical form. Here came the first work with things that have weight: clothing, stores, logistics. The discipline of judgment trained in the intangible world was brought into a physical world where many decisions cannot be undone.

The focus was not the merits of individual initiatives, but the structure in which an initiative would be decided and executed: the organization’s design itself. The defining choice came in February 2019. The Advertising Department, taken in only the previous year, was returned to PR; the IT Systems Department came in instead. After securing the visible front, it was let go in order to prioritize the less visible base of core-system integration and internal digital literacy. What to take on and what to return: the lines of the organization chart were redrawn directly each time. Keeping an organization large is not the purpose. It is only a means to build a structure through which decisions can pass.

The weight of deciding is known as personal management responsibility. That makes it possible to distinguish, for a client’s management team, whether something should be decided now or may be deferred — not as a commentator, but with the sense of a participant.

How this relationship ended became how the next stage began.

Leaving the executive-officer role did not end the relationship. Continuing it from outside was discussed at the time of departure, and became reality. The path chosen was to remain involved externally after the employment relationship ended.

Independence did not follow from leaving an executive role. The choice was to continue externally a relationship in which value had been delivered as an executive. That experience became the prototype for founding Yamazaki Office in 2020: a way of working that is neither a hired executive nor a fully outsourced vendor. “Sit beside management” was not a concept made afterward. Its first form came first.

As an External Executive Advisor, Yamazaki supports multiple companies in parallel directly under the CxOs of listed companies and holding companies. The work moved beyond individual technology and business decisions to designing the structure of decision-making itself. At its origin are both the experience of deciding personally as an executive and the choice to continue relationships after leaving that role.

What changed was the position. Through Stage 4, he was on the side that made the decisions. Now management decides, and the work is to build the structure in which management can decide. From participant in judgment to designer of judgment: the last step of this ascent.

The domain no longer remains one thing. Reframe conditions for stopping before an irreversible decision as an agreed line that lets a responsible person commit: approval design. Cut circular technical debate into the one question management must answer: re-examining customer strategy. Reject the instinct to add people because the team is busy, break down the numbers, and rediagnose a structural issue: organizational capacity diagnosis. The subjects differ; the work is one. Draw the line for who accepts responsibility, by what standard, and when.

That is why finding an answer for an individual case is not the center of the work. Designing a state in which the organization can decide for itself is the destination of this ascent. Records of individual decisions are collected in Evidence after anonymization.