Interview participant:
Mr. Daichi Matsunaga, Planning Group, Human Resources & Labor Management Department, Kobe Head Office, Kobe Steel, Ltd.

60% response rate to inquiries about year-end adjustments, realizing cost-effectiveness

-Please tell us about the background of the introduction of HUE chatbot and the challenges you faced at that time.

In our company's Human Resources and Labor Administration Department, until now, all inquiries from employees by phone or e-mail were handled by a person. If 10,000 employees make one inquiry every two months, that is 5,000 inquiries per month, and if we assume that each inquiry takes 10 minutes to respond to, it takes approximately 10,000 hours per year. However, this is just the tip of the iceberg. If we take into account the possibility that there are about five times this number of related man-hours that could be reduced, this means that up to 50,000 hours are being invested in this area. We thought that if we could automate this enormous amount of work with chatbots, it would be cost-effective enough.
At the time, chatbots from other companies were being used by the team in charge of hiring and training within the department, but they were not being fully utilized. We therefore selected several companies, including HUE Chatbot, and conducted a trial. As a result, HUE Chatbot outperformed us in all comparative items and was superior in terms of ease of use and operating costs, and we decided to implement it. At that point, we were already looking at the possibility of expanding the system to other departments, including the hiring and training team.

-Do you have any points that you were careful about or experienced any difficulties during the introduction period?

Around November 2022, we started using the HUE chatbot as a trial for employees so that we could respond to inquiries about year-end adjustments, which are concentrated at the end of the year. Although we did not have that much difficulty in preparing for the introduction, it was difficult to carefully examine the Q&A. The department in charge of creating the Q&A was not very familiar with the chatbot as it was being introduced for the first time, and there were hurdles in creating the questions that were expected.

-How is the response rate of HUE chatbot since its introduction?

The following year, when we analyzed the usage status until the year-end adjustment was completed on the management screen, we found that of the total 1,082 questions, 542 were "answered" and 540 were "not answered. However, there were 245 questions that were deemed "not necessary to be answered," such as questions that were only words, questions that did not make sense as sentences, and questions that had nothing to do with the year-end adjustment. We will continue to maintain the Q&A and try to increase the response rate each year.

Additional introduction of a document search function for even greater convenience

-We have heard that the HUE chatbot has evolved since the trial period in response to the functional comments and requests we received from your company. In November 2024, we added a "document search function," which allows users to register documents such as company manuals and have the AI generate responses automatically.

The document search function was implemented in preparation for this year-end adjustment, although we had several trials and gave feedback that it was inaccurate at first. It is quite different, and not only searches the contents of the registered documents, but also summarizes them, which is quite good. However, I have limited the number of documents to two. I would really like to add more, but it seems that if I register more than one document, the contents of the documents get mixed up (*). In the future, I would like to see the analysis function on the administration screen enhanced so that we can better understand the effectiveness of the document search function.

The document search function is a function that uses RAG to generate natural responses based on documents such as employment regulations and manuals uploaded or retrieved from websites. This will prevent the mixing of content.

Uses of HUE Chatbot Expanding Across Departments

-Please tell us about your current use of the system.

Applications are expanding. Starting with year-end adjustments, we have now added content for personnel systems, payroll systems, travel expense policies, personnel evaluations, and benefits. In addition, we have also set up a closed system for inquiries from personnel and labor relations staff at each business location, inquiries from recruiters, and content for employees working overseas. The usage rate of this section is low, probably due to the small number of eligible employees, and we will first make it known to the public on a regular basis so that they are aware of its existence.

-Please tell us about the future expansion of the scope of use.

The Finance and Accounting Department is planning to introduce a new expense reimbursement system under its leadership, and we are considering using HUE Chatbot to respond to inquiries related to the system. The Personnel and Labor Relations Department is responsible for inquiries related to travel expense regulations, and the Finance and Accounting Department is responsible for inquiries related to external payments and expense reimbursement, so the two departments will work together across the board. We are trying to utilize the document search function to proceed after the new system is put into service. Other than that, we often hear from mid-career hires that they do not know where to look for internal regulations, so we are considering having them utilize the HUE chatbot more.

-Do you feel that you are seeing the savings you initially anticipated?

We are estimating the effect of the reduction. For the year-end adjustment, let's assume that we were able to answer 500 questions, and that it took 5 minutes each for the employee to look up the contact information and ask the question, and for the department in charge of the question to answer the question. 500 questions x 5 minutes x 3 is the total time saved, and we have already achieved a reduction of approximately 120 hours. The annual estimated time for inquiries other than year-end adjustment related to personnel and payroll, personnel and labor affairs, overseas workers, and benefits was approximately 630 hours, so we estimate at this time that the total time savings, including year-end adjustment, will be approximately 750 hours.

The hurdle for introducing chatbots has been lowered

-What is your message to companies considering the introduction of chatbots?

The introduction of the HUE chatbot has been quite effective, at least for our company, and the hurdles for implementation are not that high. In particular, since the document search function has been implemented, I think it would be a good idea to try just searching for documents instead of creating Q&A from scratch. Once you register existing manuals, guidance materials, etc., the AI will answer your questions, so it would be a good idea to supplement the document search with a Q&A. This way, the initial introduction hurdle will be lowered. This would lower the hurdle for initial introduction, and I think it would make it much easier to respond to inquiries by itself.

*This article is current as of May 2025