日本語 
Hakky Inc., a company that develops AI products and supports data utilization with the mission of “adding value to products through data,” has developed “Akari AI (beta version),” a conversational AI service specialized in asset formation consultations, planned by real estate tech company TAPP Inc.
Hakky provided consistent support for the development of this service, from technology selection to RAG (Search Augmentation and Generation) construction, AI Agent implementation, UI/UX design, and implementation, contributing to the realization of an AI product that combines “accuracy” in the financial field with “friendliness” that is user-friendly.
Background of development: Removing barriers that prevent beginner investors from taking their first steps
With approximately 60% of Japanese people said to have no investment experience (*1), TAPP aimed to build an infrastructure that would allow anyone to easily seek advice, in order to remove the barrier to entry that is “anxiety due to lack of knowledge.” However, initial prototype testing highlighted an issue where users would give up on consulting at the early stages of the conversation because they “didn’t know what to ask about”.
Therefore, rather than simply developing conversational AI, we worked with TAPP to fundamentally redesign the UX based on user psychology.
Specifically, the dialogue mode can be switched to suit the user’s condition, such as providing a conversation starter for users with vague anxieties, or a visual simulation for users looking for results.By using a technological approach that utilizes cutting-edge generative AI, we have succeeded in creating a “conversational experience that allows you to consult naturally.”
Also Read: Pling Reports Japan Growth and New Features
Hakky’s technical approach to “Akari AI”
[Accuracy x Humanity] Reproducing the “personality” of experts using the latest generative AI and RAG
In the sensitive field of asset formation, providing false information (hallucination) can be fatal. Hakky has used its unique database construction technology to build a RAG system that uses AI to reference TAPP’s extensive seminar experience and specialized knowledge.
Furthermore, through advanced prompt engineering and the construction of an AI agent, we were able to establish a character called “Akari” that empathizes with the user’s concerns in an emotional way, rather than the mechanical responses that are common in financial AI.
[UX Improvement] “Mode Switching” and “Dynamic UI” to Eliminate User Confusion
To answer users’ concerns about not knowing what to ask about investments, the app has implemented a “Planning Mode” that allows users to have conversations with the goal of planning mid- to long-term asset formation. It also includes a “Quick Reply” function that proactively suggests the next question the user should ask Akari AI.
Planning mode : Extracts user attributes from the dialogue and estimates future asset trends. Developed an “Asset Strategy Report” function that generates and visualizes the results not only as text but also as individual HTML pages.
Quick reply function : We have implemented a “quick reply function” that uses AI to read the context of the conversation and suggest the next question to ask, ensuring that even passive users can keep the conversation going.
[Preventing abandonment] “Controlling response depth” creates stress-free conversations
Some users prefer to have a thorough consultation, while others prefer a quick and easy consultation first.
To meet both of these needs, Hakky has developed logic that dynamically adjusts the speed at which answers are generated and the depth of information provided. It uses two different back-end processes: a “light mode” that prioritizes response when you want a quick conversation, and a “deep mode” that requires more time to think when you need detailed advice. This optimizes the “conversation tempo,” which is important in web customer service.
SOURCE: PRTimes
English