JR West Customer Relations and ELYZA Achieve 50% Reduction in After-Call Work Time by Establishing Generative AI Summarization in Operations for 100 People

JR West Customer Relations Co., Ltd. (JWCR), which operates the "JR West Customer Center" (President: Eriko Tsutsumi), and ELYZA Inc. (ELYZA), which is advancing the social implementation of Large Language Models (LLMs) (President: Yuya Soneoka), announce that they have succeeded in reducing the average after-call work (ACW) by approximately 50% through operational improvements and continuous refinement by utilizing generative AI for the task of creating call handling records for customer "opinions and requests" received at the "JR West Customer Center."

About JWCR and ELYZA's Initiative
Since 2022, JWCR and ELYZA have jointly undertaken a DX project utilizing generative AI, aiming to improve the quality of customer support operations and reduce operational burden. As part of this initiative, they are summarizing and textifying all call handling records for inquiries received via phone and email at the "JR West Customer Center" operated by JWCR, with generative AI performing this task.

For details, please see the press release below:
https://prtimes.jp/main/html/rd/p/000000035.000047565.html

Since the introduction of generative AI for inquiry content summarization operations in 2023, both companies have been working on functional and operational improvements to enhance quality and reduce the after-call work time for operators.

About the Initiative's Results
The area where results have been achieved this time is classified under "opinions and requests" among inquiries, characterized by high complexity and difficulty in summarization. These often involve a series of statements within a single conversation, such as "opinions on services or operations," "requests for improvement," "explanations of past circumstances," "confirmation of facts," and "desired future actions," requiring multiple pieces of information to be retained in the summary. Furthermore, in cases requiring subsequent confirmation or callbacks, it is crucial to be able to trace the points to be confirmed and the prerequisites to be handed over to relevant departments from the summary. Previously, there was a challenge in organizing confirmation items and prerequisites, leading to omissions or inconsistent levels of detail in the summaries depending on the operator.

For these highly difficult-to-summarize inquiries, the monthly average after-call work time, which was 12 minutes and 47 seconds in August 2023 when generative AI was introduced for email and phone summarization, was reduced to 6 minutes and 23 seconds by December 2025, achieving a reduction of approximately 50%.

The average of 6 minutes and 23 seconds was also maintained for the three months from November 2025 to January 2026, indicating stable operation.

At the time of the introduction announcement in 2023, it was in the demonstration experiment phase, with verification conducted at some desks, showing a reduction of about 21% in time. As a result of continuous improvements by both companies since then, this has moved beyond a mere demonstration framework to achieve significant operational efficiency through practical application in organized, collective operations for an organization of 100 people.

Factors for the Achievement
Since the introduction in 2023, JWCR has achieved reductions in after-call work time by establishing rule books, conducting training, and updating AI models for improved accuracy. The main improvement measures implemented include:

JWCR's Measures
① Daily Enlightenment Activities and Revision of Operational Rules
- Explaining operational changes to small internal teams and educating them on the ideal state for a contact center and VoC analysis.
- Guiding operators to practice utterance training and call repetition to ensure AI can process effectively.
- Creating incentives for daily practice by using this as an evaluation criterion.
- Shifting operator check tasks to supervisors based on operational changes, ensuring clear role division.
- Operators: Conduct factual verification of AI-summarized content.
- Supervisors (Checkers): Review escalated cases.

② Establishment and Two Revisions of After-Call Work & Summarization Rulebook

③ Regular Operator Training Sessions

ELYZA's Measures
① Continuous Improvement of Summarization Accuracy through Three Generative AI Model Updates

② Addition and Modification of Notation Rules for Characters and Symbols

③ Improvement of Summarization Result Comprehensiveness
- Modifications made to produce summarization results with comprehensive and appropriate granularity based on supervisor check requirements. This enabled comprehensive summarization of conversation content under conditions aligned with actual operations, significantly reducing the burden of supervisor checks.

Through these measures, by confirming issues and improving business flows within the application to resolve them, and through regular sharing of operational status, close communication between the two companies has enabled continuous reduction of after-call work time, leading to the current achievement of a 50% reduction.

Comments from Both Companies Regarding This Initiative
▼ Takatoshi Iwasaki, Director and General Manager of the 1st Operations Division, JR West Customer Relations Co., Ltd.
In contact centers, the summarization of customer conversations is an essential task that requires accurate recording, but it is also time-consuming and labor-intensive for operators. Especially in our case, with a wide range of guidance topics and varying key points to record depending on the inquiry, highly advanced summarization skills have been required to accurately grasp the content of the interaction and concisely compile important points.

By utilizing ELYZA's generative AI technology to summarize and textify interaction histories based on transcribed call data, we have achieved operational efficiency while retaining summaries that cover necessary information comprehensively. As a result, this has led to the significant achievement of reducing the average after-call work time (ACW) by approximately 50%. Furthermore, by standardizing summary quality and accumulating interaction histories in an organized manner, both the quality and quantity of data usable as Voice of the Customer (VoC) have improved, expanding the potential for service improvement.

We aim to further advance the sophistication of our contact center operations through AI utilization and contribute to providing an even better customer experience.

▼ Daiki Matsuura, Manager, Solutions Business Division, ELYZA Inc.
The introduction of generative AI applications itself is merely the starting line. To achieve and sustain results through organized, collective use in actual operations, not just during the demonstration phase, it is essential to continuously refine both the on-site operations and the AI models.

We believe that the achievement of a 50% reduction in ACW was realized through the combined efforts of JWCR's diligent accumulation of daily enlightenment activities, rulebook development, and operator training, along with our repeated adjustments to prompts and improvements in the comprehensiveness of output content. The fact that stable results have been achieved in areas with high summarization difficulty, such as "opinions and requests," is a testament to the continuous improvements made from both operational flow and AI perspectives, based on shared challenges.

We will continue to work closely with JWCR to contribute to the further advancement of contact center operations.