Summary of Research Findings

Generative AI, which has been released one after another since 2022, has attracted attention around the world, including Japan, where some companies have begun to implement it. From June to August 2024, we conducted a survey on the use of generative AI among 453 private Japanese companies engaged in process manufacturing, processing and assembly manufacturing, services, distribution, and finance. Of the 453 companies, 449 companies responded to our survey, of which 4.0 percent used generative AI company-wide, and 21.8 percentv used generative AI in some departments. 

In the previous survey, conducted from June to August 2023, 1.3 percent of the 538 companies described above were using generative AI corporate-wide and 8.6 percent were using it in some departments, for a total of only 9.9 percent of the 538 companies were using generative AI. As a result of the 2024 survey, the use of generative AI corporate-wide increased by 2.7 percentage points, and the use of generative AI in some departments increased by 13.2 percentage points. Overall, there was an increase of 15.9 percentage points in just one year. It also showed that 25.8 percent of companies started using generative AI (a total of 4.0 percent of company-wide use and 21.8 percent of use in some departments), meaning that generative AI has become a prevalent technology in companies.

Noteworthy Topics

As Many Generative AI Implemented Companies Felt Below Expectations, Despite Some Effective Impacts, Expertise and Mechanism to Measure Impact In Need for Expanded Use

Of the 115 companies that deployed generative AI across the company or in some departments out of the 449 companies that responded to our survey, those that said that the impact of generative AI was far beyond expectations accounted for only 0.9 percent. Those with an overall impact in line with expectations accounted for 18.3 percent, while those with some impact but below expectations accounted for 35.9 percent. While the total number of companies that felt the impact of generative AI was 54.9 percent, the impact of implementation seems to vary by company, as there were still many companies that had no expected impact or had not measured the impact.

The reason for this is the fact that ChatGPT and other general-purpose services are the mainstream of the deployment. These services can answer general questions comprehensively, but they tend not to be able to answer company-specific work questions, which tends to limit the use of generative AI to certain areas. To make use of generative AI in specialized areas, it must be tuned using internal data and designed to fit the domain expert, which requires technological expertise and additional cost. 

In addition, the evaluation of generative AI depends on subjective feelings. When the goal is vague, such as “improving work efficiency”, it is difficult to measure the impact in numerical terms. In order for companies to continue investing in generative AI, they need to be able to demonstrate a satisfactory impact. Therefore, a mechanism for verifying the benefits or impact based on qualitative indicators is needed to maintain the prevalence of generative AI.

Research Outline