Nitori Announces Project Results for "Contact Center Reform through Collaboration between Humans and Generative AI"

Givery Inc. (Head Office: Shibuya-ku, Tokyo; President and Representative Director: Takashi Ide; hereinafter referred to as Givery), together with Nitori Co., Ltd. (Head Office: Sapporo, Hokkaido; Chairman and President, Representative Director: Akio Nitori; hereinafter referred to as Nitori), has been promoting a "Contact Center Reform Project*1" utilizing generative AI since May 2024.

We are pleased to announce that as a result of the initiatives undertaken in this project, we have achieved a reduction in operational workload equivalent to approximately 30 people, and by reallocating the freed-up resources to high-value-added tasks such as remote customer service, we have simultaneously improved customer convenience and operational productivity.

*1 November 28, 2024, "Givery Promotes "Contact Center" Project Utilizing Generative AI with Nitori." (https://givery.co.jp/news/20241128_1/)

About Nitori's Contact Center Reform Project

Nitori, in collaboration with Givery, has been promoting the "Contact Center Reform Project*1" since 2024 with the aim of maximizing customer satisfaction by restructuring the roles of humans and systems.

This project aims to build a future-oriented customer service system that can resolve customer inquiries 24 hours a day, 365 days a year (even during times when human support is unavailable) by utilizing generative AI technology and a next-generation knowledge base in the contact center. Givery is responsible for the conceptual design, infrastructure construction, and operational design in this project.

Design of Nitori's "Generative AI Chatbot for Customers"

Givery has built a "knowledge base" that centrally manages Q&A and product databases serving as reference data for external (customer-facing) FAQs, chatbots, visual IVR*2, emails, and inquiry form responses, as well as internal (employee-facing) Q&A and operational rules. This has allowed for the organization and integration of previously dispersed knowledge, enabling responses across multiple channels to be updated simultaneously.

Furthermore, based on the organized knowledge, a chatbot has been developed that uses generative AI to respond to customer inquiries on the FAQ site and each product page of Nitori Net. To improve response accuracy, knowledge is hierarchically organized and tagged, and the knowledge used in RAG is appropriately controlled.

Additionally, the AI "pre-sorts" inquiry content to distinguish between inquiries that can be handled by generative AI and those requiring human intervention, thereby ensuring guidance to the appropriate channels and enabling escalations.

Currently, the generative AI chatbot handles inquiries related to "Nitori Net (Purchases/Services)" and "Membership/Login Related," covering two areas: "1 Long-tail inquiries*3" and "2 Product inquiries."

Through these initiatives, we are promoting improved customer self-service rates and increased efficiency in customer support operations while continuously maintaining and improving our knowledge base.

【Handling of "Long-tail Inquiries"】

<Challenge> Long-tail inquiries, while not frequent individually, constitute a significant number for the contact center as a whole. The challenge has been the time required to search for information needed for responses and to confirm with administrators.

<Approach to the Challenge>

Deployment of a generative AI chatbot.

Extraction of unanswered questions and improvement/expansion of knowledge.

Customers can easily ask questions via chatbot that they might hesitate to ask via phone or email.

This clarifies minor doubts or uncertainties customers may have. For Nitori, once the knowledge (Q&A and operational rules) for these doubts is established, the chatbot's generative AI responses significantly reduce the time required.

<Chatbot Design>

Standard inquiries: Rule-based selection for common questions (Q&A) is prepared.

Answers frequent questions without requiring customers to type extensively.

Long-tail inquiries: Customers can choose whether to "ask the generative AI" or "contact a human operator."

While responses from human operators may involve waiting times depending on the time of day, generative AI provides answers 24/7 without any waiting.

Generative AI Responses

⚪︎ Control of RAG target text: By limiting the knowledge reference folders, the risk of hallucination*4 is suppressed.

⚪︎ Pre-sorting by AI: For inquiries such as complaint handling or product recommendations, where generative AI responses are not appropriate, the AI pre-determines and guides to the correct channel.

Analysis of response logs: By analyzing self-resolution rates for each inquiry route, areas for improvement become clear.

Customer questions are used to identify unclear points, not only for knowledge improvement but also for UI/UX improvements on Nitori Net.

【Handling of "Product Inquiries"】

<Challenge> Nitori Net handles over 160,000 products, leading to numerous inquiries about them. Furthermore, products are frequently updated, making it impractical to create individual Q&As and knowledge bases for every product. Consequently, most inquiries could only be handled by human operators.

<Approach to the Challenge>

Deployment of a generative AI chatbot for each product page.

Development of data linkage with Nitori Net and product databases.

Extraction of unanswered questions and improvement/expansion of knowledge.

<Chatbot Design>

Product identification: Automatically retrieves the product code displayed on Nitori Net.

Standard inquiries: Rule-based selection for common questions (Q&A) is prepared.

(Responses are provided from the product master / e.g., shipping fees, assembly requirements, etc.)

Generative AI Responses:

⚪︎ Control of RAG target text: By referencing knowledge in the order of product master → category knowledge → common knowledge, it broadly covers specific products, categories, and general inquiries while suppressing the risk of hallucination.

⚪︎ Pre-sorting by AI: For inquiries such as product recommendations, where generative AI responses are not appropriate, the AI pre-determines and guides to the correct channel.

⚪︎ Analysis of response logs: Unclear points are identified from customer questions, and missing information or explanations are pinpointed.

This information is reflected in the product description sections of Nitori Net and also added to the knowledge base referenced by store employees, enabling its use in answering inquiries at physical stores.

*2 Visual IVR: A system that displays choices on a website according to the inquiry content and guides users to appropriate information or contact points. *3 Long-tail inquiries: Diverse inquiries that occur infrequently and cannot be fully addressed by FAQs. *4 Hallucination: The phenomenon where generative AI produces inaccurate responses not based on facts.

Project Achievements

Through these initiatives, the contact center has achieved the following results:

10% year-on-year reduction in human-handled inquiries via phone, email, and live chat.

Achieved 24/7 inquiry support through the generative AI chatbot, enabling customers to self-resolve issues, including during nighttime hours.

Reduced operational workload by approximately 30 person-equivalents, reallocating the created resources to high-value-added customer service tasks such as remote support.

Improved Nitori Net's UI/UX based on free-text customer questions.

Future Outlook

Givery and Nitori will continue to pursue systems that offer "greater convenience and peace of mind" by evolving the optimal collaboration between automated and human support using generative AI.

Furthermore, by leveraging knowledge assets using AI agents, we will promote DX (Digital Transformation) and AX (AI Transformation) not only for external customers but also for internal employees, and will continue to expand functions and improve operations to enable employees to focus more on high-value-added customer service.

From the Project Managers

Hiromichi Iwatani, Manager, Sapporo Call Center, Nitori Co., Ltd. Nitori has taken another step closer to providing a 24/7 support system to deliver value beyond expectations to our customers. We recognized the need to tackle generative AI responses ourselves, given the risk of misleading customers, for example, through AI summaries in search engines. While simple inquiries that involve looking up and answering can now be handled by the system, the proportion of work requiring human intervention for high-value-added tasks has increased. In the next step, we would like support in creating mechanisms to streamline the tasks that humans should handle.

Masaaki Yoshida, Director, CMO, and COO of Marketing DX Department, Givery Inc. "At Givery, we believe that by combining generative AI and knowledge management, we can achieve both improved employee productivity and customer satisfaction, enabling customer support where humans and AI leverage their respective strengths. We will continue to create new customer experiences by maximizing the use of digital technology together with Nitori."

Company Overview

【About Nitori Co., Ltd.】 Company Name: Nitori Co., Ltd. Representative: Akio Nitori, Chairman and President, Representative Director Head Office Location: Sapporo Head Office, 7-jo 1-chome, Kita 7-jo Nishi, Kita-ku, Sapporo, 001-0907 Established: March 1972 Business Activities: Planning, development, and sales of furniture and interior products (home furnishing products), corporate business, renovation business, etc. URL: https://www.nitori.co.jp/

【About Givery Inc.】 Company Name: Givery Inc. Representative: Takashi Ide, President and Representative Director Head Office Location: Teito Shibuya Building 8F, 15-13 Nanpeidai-cho, Shibuya-ku, Tokyo, 150-0036 Established: April 28, 2009 Capital: 50,000,000 yen Business Activities: Consulting/System Development Business, AX (AI Transformation) Business, Marketing DX Business, Human Capital Intelligence Business, Cybersecurity Business URL: https://givery.co.jp/