Bellsystem24 Launches 'Sherpy' AI Chat Navigator to Accurately Handle Complex Inquiries via 'Hybrid RAG'

Bellsystem24, Inc. (Headquarters: Minato-ku, Tokyo; President and CEO: Hiroshi Kajiwara; hereinafter 'the Company') has announced the launch of 'Sherpy' (tentative name), an AI chat navigator designed to accurately respond to complex inquiries.

'Sherpy' leverages 'Hybrid RAG' (Retrieval-Augmented Generation), a technology that enhances response accuracy, allowing the AI to autonomously diagnose user intent through a chat interface and derive precise answers from stored knowledge. Through integration with call center interactions, the AI is continuously refined, enabling it to handle complex queries and achieving a level of accuracy unattainable by standalone AI systems.

The Company is rolling out two models: 'Sherpy for Operator' (tentative name) to assist operators during customer interactions, and 'Sherpy for Customer' (tentative name) to support customer self-service. These models aim to simultaneously improve operator efficiency and enhance customer experience (CX) in contact centers.

This solution is the second release under the 'Hybrid Operation Loop' banner, an automated generative AI suite for contact centers. Currently, several companies, including major life insurance and energy providers, are conducting proof-of-concept (PoC) trials, with full implementation to follow sequentially.

### Background Through the 'Generative AI Co-Creation Lab.', a program for building hybrid contact centers using both AI and humans, Bellsystem24 received feedback from clients stating that existing AI chatbots often failed to meet accuracy expectations for difficult inquiries, forcing users to rely on human intervention.

Automating high-precision responses requires consolidating scattered knowledge like FAQs, manuals, and operator expertise. However, this consolidation is often prohibitively costly and time-consuming. Furthermore, standard RAG models based on keyword similarity often struggle with inquiries involving complex user intents.

To address these issues, Bellsystem24 first developed 'Knowledge Generator' to automatically create high-quality, KCS-compliant knowledge from voice data. The second solution, 'Sherpy', utilizes this knowledge base through 'Hybrid RAG'—a technology that derives answers based on the relationships between multiple pieces of information—to automate complex inquiry handling.

### Features of Sherpy The name 'Sherpy' is derived from 'Sherpa,' the experienced guides who safely lead climbers to mountain summits. Sherpy acts as a guide, leading users to the 'summit' of a correct answer using its high-quality knowledge base.

Aligned with the concept of 'Human-in-the-Loop,' humans and AI collaborate to train and evolve the system, much like an instructor trains an operator. Sherpy incorporates Bellsystem24's extensive operational expertise across industries such as insurance, energy, telecommunications, and finance. This deep understanding of practical challenges—such as identifying high-difficulty queries and operator decision points—enables the creation of an AI navigator that is truly effective in the field.

Two models have been developed: - **Sherpy for Operator (tentative name)**: Assists operators by providing accurate AI-generated answers in response to text queries during customer service.