BELLSYSTEM24, Inc. (headquartered in Minato-ku, Tokyo; Hiroshi Kajiwara, President, Representative Director; hereinafter "BELLSYSTEM24") announces the release of "Sherpy (tentative name)," an AI chat navigator capable of accurately responding to complex queries.
"Sherpy (tentative name)" utilizes "Hybrid RAG^*1" technology for improving response accuracy. The AI autonomously detects user intent via chat and draws on accumulated knowledge to deduce an accurate response. Through dealing with customers at contact centers, the AI grows stronger until it can deal with the most complex of queries, resulting in a level of response accuracy beyond the reach of AI alone.
BELLSYSTEM24 will offer two models: "Sherpy for Operator, (tentative name)" designed to support operators in dealing with customers, and "Sherpy for Customer (tentative name)," designed to support customers in resolving queries themselves, with the dual aim of increasing the operational efficiency of contact center operators and improving the customer experience (CX).
This is the second solution derived from the contact center generative AI automation solution "Hybrid Operation Loop^*2." BELLSYSTEM24 is currently conducting demonstration experiments at several companies including a major life insurance company and a major energy company, and will begin implementation sequentially.

URL：https://www.solution.bell24.co.jp/ja/solution/data/hybrid-operation-loop/

■ Background

BELLSYSTEM24 has received a great deal of feedback from client companies through the activities of the GenAI Co-Creation Lab. program. This is a program with the involvement of user companies that enables participating companies to share examples of best practices with a view to constructing a hybrid contact center leveraging both AI and human resources to be operated by BELLSYSTEM24. Some clients reported that the introduction of an AI Chatbot did not result in the expected level of accuracy and ultimately led to reliance on human resources to deal with complex queries.
The automation of highly accurate responses using generative AI requires the aggregation of knowledge scattered throughout the contact center, such as FAQs, manuals, and the tacit knowledge of operators. However, the aggregation of this knowledge is enormously costly and time-consuming, and in reality, many companies are not getting to grips with it. Another issue is that even AI agents using RAG (Retrieval-Augmented Generation) to improve response accuracy perform searches based on similarity between queries and responses and are still incapable of dealing with complex, intent-based queries.
To resolve these issues, BELLSYSTEM24 developed "Knowledge Generator" as a first solution derived from the "Hybrid Operation Loop." This solution automatically generates KCS^*3-compliant high-quality knowledge from voice call data, leading to the development of a knowledge base that generative AI can refer to. "Sherpy (tentative name)," the second solution derived from it, uses "Hybrid RAG" search technology that arrives at a response by retrieving data from this knowledge base and evaluating the contextual relevance of multiple data sources to achieve the automation of complex queries.

■ Features of "Sherpy (tentative name)"

The name "Sherpy (tentative name)" comes from the word "Sherpa" used to refer to experienced guides who lead mountain climbers safely to the summits of the world's most inaccessible mountains. "Sherpy (tentative name)" is an AI chat navigator that draws on a high-quality knowledge base to tackle even the most complex queries, reliably leading to the "summit," that is, an accurate response.
In the same way that training officers develop operators through training and OJT before they start work, the "Hybrid Operation Loop" will also train AI in collaboration with humans based on the "Human-in-the-Loop" concept to develop AI capable of giving highly accurate responses. "Sherpy (tentative name)" was developed based on operational knowledge accumulated by BELLSYSTEM24 through contact center operations across a wide range of industries including the insurance, energy, telecoms and financial sectors. BELLSYSTEM24 is capable of developing AI and releasing an AI chat navigator precisely that is actually usable in the field because of its familiarity with the practicalities of contact center operations, including the kind of queries that are highly complex and the situations in which contact center operators are unsure. BELLSYSTEM24 developed two models of this new AI chat navigator as follows.

Sherpy for Operator (tentative name) (operator response support model)

When an operator dealing with a customer asks a question via text, "Sherpy (tentative name)" uses Hybrid RAG to detect the intent of the question and quickly retrieve accurate potential responses and relevant information from the knowledge base. Enhancement of the AI chat navigator's ability to deal with highly complex queries will help improve response quality and shorten query handling time. It will also help shorten the time it takes to train new operators and reduce employee turnover.

Sherpy for Customer (tentative name) (customer self-solving support model)

The AI chat navigator is installed on the company's website. "Sherpy (tentative name)" draws on the knowledge base to respond to customer queries. With existing FAQ and scenario-based chatbots, customers with complex queries tended to be put through to a human operator in the end but, with "Sherpy (tentative name)," even the most complex queries can be dealt with 24 hours a day, 365 days a year. By enabling customers to resolve queries themselves, "Sherpy (tentative name)" will deliver improvement in customer satisfaction and greater operational efficiency through reduction in calls put through to a contact center.

■ Future Outlook

BELLSYSTEM24 will make functional enhancements to "Sherpy (tentative name)" to further increase response accuracy and also build up a track record with major companies, aiming to increase the number of companies using it to 50 by 2028. Additionally, the development of an AI voice agent to automate response support is now under development as a third solution. BELLSYSTEM24 aims to realize "next-generation contact centers" in anticipation of fully-automated customer service.

■ About related solution "Knowledge Generator"

Knowledge Generator, which is the first solution derived from the "Hybrid Operation Loop," has been provided as part of BELLSYSTEM24's offering of knowledge management consulting services. Now its system environment is in place, BELLSYSTEM24 will also phase it in as SaaS. This solution massively shortens the enormously time-consuming process of operators developing KCS-compliant knowledge from voice call data at contact centers, making the process far more efficient, and contributes to the generation of high-quality knowledge.