TRUSTBANK Leverages AI Agents for Personalized Furusato Nozei Gift Recommendations

AI-Powered Gift Discovery in Japan’s Hometown Tax Program

TRUSTBANK partnered with Recursive to launch Choice AI, a conversational AI feature within its Furusato Choice platform, designed to simplify the selection of “thank-you gifts” associated with Japan’s Furusato Nozei program. The platform lists approximately 760,000 gifts, creating a challenge for users to find relevant options.

Why This Matters

The Furusato Nozei program, while beneficial for regional economies, presents a significant user experience challenge: information overload. Traditional e-commerce recommendation systems struggle with the unique context of this program, where users aren’t necessarily seeking specific products but rather the best use of their donation limit. Failing to address this complexity leads to user frustration and potentially reduces participation in a program intended to support local communities – representing a loss of potentially billions of yen in regional funding.

Key Insights

• Multiagent Architecture: Choice AI employs a routing model delegating tasks to specialized agents (Search, Recommendation, Greeting) for efficient task handling.
• RAG Systems for Context: Recursive built a Retrieval-Augmented Generation (RAG) system to ground the AI’s responses in the thank-you gift database.
• Dynamic Prompting: The system adjusts prompts based on user type (new vs. returning) to personalize the interaction flow.

Practical Applications

• Use Case: TRUSTBANK utilizes Choice AI to guide users through the vast Furusato Nozei gift catalog, increasing conversion rates by understanding vague user needs.
• Pitfall: Relying solely on keyword search in a complex donation program like Furusato Nozei can lead to irrelevant results and user abandonment due to the program’s unique context.

References:

• https://openai.com/index/trustbank/