AI Agents Leverage Japanese Yen Stablecoin JPYC to 'Vibe Code' Behavioral Change in Health Management

Sense It Smart Inc. (Headquarters: Chiyoda-ku, Tokyo; CEO: Hiroshi Tanimoto; hereinafter "the Company") announces the launch of recruitment for joint Proof-of-Concept (PoC) partners to test its "Health Management AI Agent Package" (hereinafter "the Package"). This initiative will verify the feasibility of using Japan's yen-pegged stablecoin "JPYC" (an electronic payment instrument under the Fund Settlement Law) for conditional reward distribution.

In this PoC, corporate health management officers (hereinafter "officers") will issue natural language instructions to AI agents (such as Claude Code or Codex), testing whether the system can manage the entire process—including challenge design, participant invitations, achievement verification, JPYC distribution, and report generation—with human approval at key stages.

While typical walking apps focus on step tracking, rankings, and point accumulation, the Package aims to become an "execution platform for behavioral change in physical activity"—one that AI agents can invoke externally to handle implementation, achievement assessment, reward distribution, and iterative improvement.

AI agents can generate plans and reports, but cannot act in the "real world" independently. Today, AI is beginning to control domain-specific "execution platforms" through MCP (Model Context Protocol), a standardized method for AI to interact with external systems. Examples include GitHub for development, Stripe for payments, Slack for internal communication, freee or Money Forward for accounting, and Sansan for sales. Sense It Smart’s "Exercise Supplement®" serves as the execution platform for health management and behavioral change through physical activity.

The Company is bringing the concept of "vibe coding"—a methodology spreading in software development—to the operation of health management, using this execution platform as its foundation.

"Vibe coding" refers to the ability to instruct AI in everyday language (e.g., "build this") to implement and run software, make corrections (e.g., "fix this"), and iterate rapidly. Even non-engineers can now shape their desired software. However, this has so far been limited to the "digital realm." The Package extends this concept beyond the screen—into the real world—where employees take actual physical actions and money moves in reality.

※ Employee participation is voluntary. The AI does not control employees themselves, but rather the process of designing, executing, and improving health initiatives.

The goal is for AI agents to manage the entire lifecycle of physical activity programs—from planning and execution to reporting—based solely on natural language instructions from officers. Human approval is required at critical points involving fund transfers and physical exertion. Upon achievement verification, rewards are automatically distributed in JPYC via smart contracts (programs that execute automatically based on predefined rules). This is the essence of "vibe coding" for health management—enabling small teams to manage health programs effectively amid structural labor shortages.

Previously, AI and health initiatives were limited to advice and environmental support (e.g., budgeting, tool provisioning, usage tracking, and reporting), often failing to drive actual behavioral change. The Package goes further: it operationalizes real employee actions and distributes rewards (in JPYC) upon achievement (currently implemented for walking).

The core execution platform, "Exercise Supplement," is already operational as a service. While AI can handle planning, announcements, and reporting, delivering these components in a fully integrated, optimized manner remains a future challenge. Currently functional components include challenge creation, step data submission, achievement verification, and JPYC reward distribution. The overall quality of operations—including announcements, follow-ups, and reporting—will be evaluated through this PoC.

The Package follows an "AI-operated, human-approved" design: AI manages operations while humans approve and sign off at critical points involving fund transfers and safety. This architecture aims to make "vibe coding" robust enough for real-world health management deployment (details on safety design are provided later).

The overall flow of the Package is as follows: from instruction to planning, execution, distribution, and reporting—with human approval at key stages.

Example of Natural Language Instruction:

Officers need only issue natural language instructions to the AI agent. The same method applies not only to launching initiatives but also to mid-campaign follow-ups and designing the next challenge based on results.

"It seems our employees’ activity levels have decreased recently. I’d like to implement a walking initiative for 100 employees within a budget of 100,000 JPY. Please make it manageable and handle everything from announcement to follow-up and report generation."

Design to Reduce Officer Workload and Company-Wide Costs

The Package aims to reduce human dependency in operations, thereby decreasing both individual officer workload and overall corporate program costs.

Procurement Effort: Replacing vendor requirement alignment, internal coordination, and revision cycles → Single natural language instruction

Operational Labor: Replacing manual planning, announcements, participant management, data aggregation, and reporting → AI handles operations; officers focus on approvals

Program Costs: Replacing piecemeal external contracting → Consolidating into an internal platform to reduce TCO (Total Cost of Ownership) across multiple initiatives

※ Reduction levels vary by environment. Quantitative effects will be measured and verified during the joint PoC (the ability to measure impact within one’s own environment is itself a benefit).

Why JPYC Now? AI-Ready, Programmable Reward Distribution

In an era where AI agents manage program operations and execution, a form of "money" that AI can directly control—without human intervention—is essential. Unlike traditional payment methods requiring manual processing, programmable stablecoins allow AI to automatically execute reward distribution from achievement verification onward. Stablecoins are highly compatible with AI agents.

The Japanese yen-pegged stablecoin "JPYC," used as the reward source, is compatible with multiple blockchains. For this PoC, Polygon or Kaia will be used. JPYC is not used to "motivate with money," but because its distribution rules—such as recipients and ratios upon achievement or non-achievement—can be pre-programmed. By configuring these rules, the "reason to walk" can be personalized for each individual. JPYC is a legally recognized electronic payment instrument under the Fund Settlement Law, with a 1:1 peg to the Japanese yen (1 JPY = 1 JPYC). It is not a speculative cryptocurrency but a programmable reward distribution rail. In this PoC, such conditional reward distributions will be operated with individual consent and human approval, verifying whether health management initiatives can be iteratively improved by learning from each cycle’s results.

Another advantage of using programmable stablecoins is the ability to automatically execute conditional, multi-party distributions based on achievement or non-achievement—something difficult with traditional points or gift vouchers—and do so in a tamper-resistant manner.

This initiative aligns with the full-scale launch of the Japanese yen stablecoin "JPYC" (October 2025; cumulative issuance on the JPYC EX redemption platform [money transfer operator type] exceeded 3 billion JPY by May 30, 2026, with 19,000 accounts opened, per JPYC Co., Ltd.) and the LDP’s proposal "Next-Generation AI and On-Chain Finance Vision" (May 2026), which share the same vision for programmable yen stablecoin utilization.