HealthTech AI in Japan: Define the Intended Use Before the Pilot

HealthTech AI projects often start with a model idea. At Tokyo AI's Japanese Industry Overview session, Joshua Owoyemi's talk placed that idea among the pressures and constraints of Japan's health sector. The first design question is more basic: who will use the system, for which decision, and in what setting?

Separate the use cases before building

The talk distinguished operational uses such as scheduling and intake from assisted diagnosis and research workflows. They should not share one vague definition of success. An appointment assistant can be evaluated for correct routing, accessibility and staff time saved. A tool that informs clinical judgment needs a different evidence and review process. A research aid may be valuable even if it never produces a patient-facing output. Calling all three “medical AI” hides those differences.

Joshua also discussed the pull of an aging society and the push of new AI capabilities. Neither is a deployment plan by itself. A hospital, care provider or research group has existing records, staff routines and procurement constraints. A pilot must fit these conditions and identify the person responsible for its output. A promising demonstration that cannot be corrected within the workday may add burden rather than reduce it.

Define an intended use that can be tested

Consider a non-clinical intake assistant as an illustrative starting point. Its job might be to sort incoming requests into categories for staff review, not to decide a diagnosis. The team can specify allowed inputs, language handling, escalation for uncertainty and what staff see when they correct a category. It can then measure misrouting, time saved and the work that remains with people. If the intended use later changes, the evidence and approvals must be reviewed again.

For teams building health technology in Japan, that sequence matters more than a broad promise that AI will transform care. State the user and decision, test inside the actual workflow, and keep the boundary between operational support and clinical judgment explicit. This article discusses product and engineering design, not medical advice or regulatory approval.

Tokyo AI is a community of builders, researchers and business leaders in Japan, run by Foundry Labs. See the event page.

Materials from the speakers’ talks are available upon request. Please use our contact form and include the event title and speaker name.

Speakers

• Joshua Owoyemi, CEO, ModAstera.

All articles