I Help Organizations Close the Gap Between AI Ambition and Operational Reality

Martin Zialcita, also known by his Japanese name 大森 淳司, was born in Japan, raised in Guam, and is now based in Honolulu, Hawaiʻi.

His work brings together practical AI implementation, marketing strategy, education, and a human-centered approach to organizational change.

I spent most of my career on the commercial side of organizations: marketing, growth, customer systems, analytics, and the long grind of digital transformation projects. That background matters more to how I work with AI than any of the AI-specific parts, for one reason: I watched a great many technology projects succeed technically and fail operationally.

The pattern was consistent. A capable system would be selected, configured, and launched. Six months later people were working around it, maintaining a parallel spreadsheet, or quietly reverting to the old process. The technology was rarely the problem. The problem was that nobody had designed for how the work actually happened, who would own the new process, or what would happen when the system was wrong.

AI has reproduced that pattern at speed. The capability is genuinely remarkable and unusually easy to demonstrate, which makes the gap between a compelling demo and a dependable workflow wider than it was with previous technology waves, not narrower. Closing that gap is the work I do.

I am not interested in being the person who tells you AI will transform your organization. I am interested in whether a specific workflow runs better on a Tuesday in six months, with your staff operating it and without me in the room.

I work with leadership teams on four connected things: finding the workflows where AI would genuinely help, connecting it to the systems they already run, designing the oversight that makes it defensible, and preparing the people who will operate it.

In practice a typical engagement starts with observation rather than recommendation. I sit with the people doing the work and document how the process really runs, including the workarounds and the informal steps that exist because a system does not do something it should. That map is usually the first genuinely useful artifact, and it is often the first time anyone has written the real process down.

From there the work is ordinary engineering and ordinary change management: prioritize by value, feasibility, and risk; integrate against real systems with real permissions; decide who reviews what; train the operators on their own cases; and instrument the result against a baseline captured beforehand.

Commercial delivery happens through AI Marketing Box, my practice. This site is where I explain the thinking and publish the frameworks.

The commercial background is directly relevant rather than incidental. Marketing and revenue operations are where most organizations first encounter the specific problems AI implementation runs into: data spread across systems that disagree with each other, processes that exist because of a tool limitation nobody remembers, and measurement that cannot survive scrutiny.

Working in CRM and analytics teaches you that the interesting question is rarely “what should we do” but “what is actually happening, and how would we know”. That habit transfers directly. It is why I ask for a baseline before deploying anything, and why I am skeptical of efficiency figures, including my own, that arrive without a stated method.

It also shapes which AI work I think is worth doing first. Lead response, intake, follow-up, and reporting are unglamorous, high-volume, and measurable. They are usually a better first implementation than something more visibly impressive.

His background spans Japan, Guam, and the United States. That is not a biographical detail I include for color. It changed how I think about technology adoption.

Culture determines the pace at which a new system is trusted, and trust determines adoption far more than capability does. In relationship-driven business environments, a system that produces one embarrassing error in front of a community partner costs more than the efficiency it was purchased to deliver. Oversight design is not bureaucratic overhead in that context; it is the precondition for being allowed to proceed at all.

The Japanese learning progression of shu, ha, ri (follow the form, adapt the form, transcend the form) is also a genuinely good description of how organizations build capability with a new technology, as opposed to how they buy tools. It is the basis of one of the frameworks I teach, and it will be published as a full page in the next phase of this site.

On Hawaiʻi specifically: the economy here is structured differently from the large mainland organizations most AI advice is written for. That is covered in detail on AI in Hawaiʻi.

Martin has taught an AI Marketing and Ethics course at the University of Hawaiʻi at Mānoa.

No academic title, department, or current appointment is claimed. See teaching for the subject matter and credentials for the evidence status.

Martin holds an MBA from the Shidler College of Business at the University of Hawaiʻi at Mānoa.

The previous version of this site carried a set of aggregate figures: efficiency gains, funding multiples, organizations served, hours saved, dollars secured. None of them appear here.

That is not because they were necessarily wrong. It is because none had a recorded baseline, measurement method, sample, or client permission on file, and a figure without those carries no information regardless of who publishes it. Each is now logged in a claim register and will only be republished with its evidence and approved wording attached.

The same standard applies to superlatives. Descriptions like “premier” or “leading” are not verifiable, so they have been retired rather than queued for evidence.

If you find something on this site you think is unsupported, the editorial policy has a correction route and I would genuinely like to hear about it.