Evaluate design through the lens of a simulated Japanese customer.



Year: 2026 | Industry: Fintech | Region: Japan |  Role: UX Strategist/Prompt Engineer | Team: UX Research | Deliverable: AI Agent
  • When offshore teams and remote workers lack the time and budget to engage with Japanese customers regularly, a significant communication gap emerges. While AI offers a potential solution, standard simulated personas often represent Japanese customers superficially, or even as caricatures. Therefore, the real challenge lies in developing an AI agent that accurately reflects the nuanced, lived experiences of a real Japanese person.

  • To address this, a dual agent approach was developed. Instead of relying on a single AI agent, the solution utilizes two distinct AI agents: one dedicated to defining the cultural and professional persona, and another specifically trained to evaluate design proposals through the lens of a Japanese person.

The persona GPT combines every day Japanese behavior, financial habits, social norms and economic context.
The design review GPT uses besides socio-economic and cultural context, the core Japanese user and its associated user behavior as main inputs.
  • Both AI agents are optimized for simplicity. A designer uploads a document or idea, and the AI agent generates a structured persona or evaluation in plain English with minimal business jargon, ready to share in a slide deck or in Slack.

  • The AI agent that generates a persona, contains:
1. Target group
2. Socio-economic and cultural context
3. Personel perspective
4. What designers should understand
5. Assumptions
6. Next steps

  • The AI agent that evaluates design proposals through the lens of a Japanese person, contains:
1. Core Japanese user
2. First reaction 
3. Biggest friction
4. Likely action
5. Assumptions
6. Next steps

  • Today, both AI agents serve as either essential starting points or pivotal review milestones in design projects. While both agents perform exactly as expected, it remains crucial to recognize that their outputs are indicative and serve as a guide, rather than a definitive replacement for real Japanese users.