85% Reduction in Man-Hours! Fujitsu's Internal Practice for Visualizing and Structuring the Thoughts of Conversation Partners

Hello. I am Asama, and I promote the use of generative AI at Fujitsu.
The scope of generative AI applications is vast, and it is known that incorporating frameworks and know-how into it can yield higher-quality output. This time, I will introduce the efforts of a team (Nakamura, Inoue, and Nogawa from the Fujitsu Design Center) that developed a new internal application (hereinafter referred to as the Interview Analysis App). They combined generative AI with a proprietary interview method developed by Fujitsu,“AIm (Appreciative & Imaginative) Interview”to enable employees to perform more advanced analysis of interview content and improve the quality and productivity of their work.

In short, the Interview Analysis App we are introducing this time is a tool thatanalyzes the vast amount of information spoken by people in meetings, interviews, etc., and concisely summarizes it into seven items.

However, since it might be difficult to visualize, I would like to first show you what the Interview Analysis App is actually like.
I used the Interview Analysis App based on the recording data from when I interviewed the development team. It is easy to use; you justcopy and paste the interview transcription results into the Interview Analysis App and press the send button.

The results were returned in about 3 minutes.

The duration of this interview was about 30 minutes, and the development members shared their "thoughts," starting with an overview of the Interview Analysis App and its development background, to the difficulties and ingenuity during development, and future prospects.

In just 3 minutes, data structured and visualized into 7 items and approximately 1,800 characterswas returned.Simply summarizing content with AI can sometimes omit parts that contain the speaker's important "thoughts".
Because the Interview Analysis App analyzes content from the perspective of an expert before summarizing it, it is impressive that the output data not only included explanations of marketing strategies, user engagement improvement, technology expansion, and global development, but alsoclearly stated the "thoughts" of the developersthat were seen throughout the conversation without omission.
Now that you understand the benefits of the Interview Analysis App, here is an explanation of the "AIm Interview," which can be called the heart of the app.

First of all, what is the "AIm Interview" developed by Fujitsu Laboratories?

The“AIm Interview”, developed and released by Fujitsu Laboratories in 2008, is an interview method (qualitative design methodology) developed for the purpose of creating business solutions deeply rooted in the customer's perspective.It is possible to seamlessly create everything from ideal images that resonate with customers to action plans after visualizing and structuring the customer's statements into the following seven items.

1. Current Situation Awareness

2. Energy Source (Driving Force)

3. Values

4. Strengths

5. Ideal Image

6. Gap and Issues between Ideal and Reality

7. Action Plan

With this method,it becomes possible to examine business solutions from a medium- to long-term perspective by drawing out diverse relationships woven by actual users and needs that the users themselves are not aware of, rather than superficial needs.
"AIm Interview" has been used in job categories such as insurance sales, bank loan and liaison work, public health nurses, university faculty, university administration, elementary school teachers, and designers, and has been used within Fujitsu to structure individual and organizational visions and to assist in sales activities.

Please tell us the features of the Interview Analysis App developed this time.

The Interview Analysis App developed this time is a tool that combines this "AIm Interview"with generative AI, which is a good match.
While the "AIm Interview" alone had bottlenecks in terms of the man-hours required to analyze and revise data content, as well as the need for experience and skills,by combining it with generative AI (large language models) trained on the know-how of experts, the AI generates reports in a few minutes that were previously created by hand, leaving only the need for revisions, andit has become possible to analyze speech content and organize text "regardless of individual skills" and with "15% of the conventional man-hours".
Since around October 2023, we have conducted over 50 verifications in 12 use cases, including internal meetings, round-table discussions, questionnaire analysis, business planning, and 1-on-1s. In addition, to ensure that employees can use the app with peace of mind, we have established terms of use and have been deploying it internally since March of the following year.
"I felt that the visualized content was more concrete and accurate than I had expected!" "I was able to proceed to concrete discussions toward turning it into a project with a customer from whom it was difficult to draw out requirements." "As a salesperson, I am very surprised by the unprecedented reaction from a customer I have been visiting for a year!"Positive opinions such as these have been received.

Conclusion

Through this interview with the developers, I felt that there is still a lot of untapped potential in combining generative AI with the expertise accumulated by humans. Depending on the response, we may continue to share practical internal use cases in the future!?
Moving forward, I would like to go beyond interview analysis and take on the challenge of combining generative AI with the knowledge Fujitsu has cultivated in areas such as management and business strategy.

Recommended Articles