Corporate News
uSonar Co., Ltd. (Headquarters: Shinjuku-ku, Tokyo; hereinafter "uSonar") is pleased to announce the release of a report based on a survey regarding the actual usage and satisfaction levels of generative AI output quality, conducted among 500 business professionals currently using generative AI in their work.
The survey report reveals that 70% of users are dissatisfied with the accuracy of AI outputs, and highlights the reality that verifying these outputs consumes an unexpected amount of time.
The findings suggest that for companies to improve productivity through generative AI, there is an increasing need to focus on high-quality data management. We invite you to download the report for further details.
Download the Detailed Survey Report
To obtain the detailed report, please use the application form on the right side of this page. You may also use the "Get Survey Report" button below (this will redirect you to the free download page for the document).
Background and Purpose of the Survey
While the use of generative AI in business is expanding, there is growing dissatisfaction with AI-generated outputs, with users noting that they "cannot be used as-is" or that "poor accuracy leads to unexpected time spent on verification." This is often attributed to a lack of "prompt engineering skills." However, we conducted this survey to explore whether the issue might also lie in the "quality of data" that the generative AI references, rather than just prompt skills.
• Quantify the "Real-World Usage" of Business Professionals Utilizing Generative AI in Their Work
• Identify the "Root Causes" of Dissatisfaction Regarding AI Response Quality and Reliability
• Explore the Relationship Between AI Performance and the Necessary Data Management Required to Leverage It Effectively
Survey Overview
• Survey Name: Survey on Generative AI Usage and Quality
• Method: Internet Survey (Cross Marketing Inc. Survey Panel)
• Target: Business Professionals Who Use Generative AI "Daily" or "Occasionally" for Work
• Valid Responses: 500
• Period: March 2026
• Conducted by: uSonar Co., Ltd.
Key Survey Results (Summary)
1) Approximately 70% have experienced dissatisfaction with the "accuracy and reliability" of generative AI (70.4%)
70.4% of respondents answered that they "frequently" or "sometimes" experience dissatisfaction regarding the accuracy and factual reliability of generative AI responses. While the use of generative AI is becoming mainstream, it is clear that concerns regarding reliability remain widespread.
2) The cause of dissatisfaction is "data quality" — 61.4% attribute it to data
When asked about the causes of dissatisfaction with AI responses, 23% cited "the AI lacks specialized or up-to-date information" and 22% cited "the possibility that the source data is incorrect," with a combined 61.4% attributing the issue to the "quality of data and information."
Meanwhile, only 12.4% cited "lack of prompt skills," suggesting that many believe the core issue lies in the data environment referenced by the AI rather than the user's skills.
3) Time spent verifying responses offsets productivity gains — 72.8% spend time verifying or correcting AI outputs
72.8% of respondents spend 5 minutes or more per instance on "error checking and verification" of information provided by AI, with some spending over 30 minutes. This highlights the reality that the benefits of AI adoption are being offset by the effort required to verify outputs.
4) The key to improving AI usage is the "data environment" — 61.6% consider it the most important factor
When asked what is most necessary to improve the level of generative AI usage in the future, "improving the accuracy and timeliness of data" (33%) and "access to a wider range of factual data" (29%) topped the list, with a combined 61.6% identifying "data improvement" as the most important factor. This result surpasses "improving prompt skills" (20%).
Implications of the Survey
• The Bottleneck for Generative AI Usage Is the "Data Environment," Not "Prompts"
This survey indicates that the primary cause of dissatisfaction with generative AI outputs is not the user's AI skills, such as prompting, but rather concerns regarding the "accuracy," "timeliness," and "expertise" of the information data referenced by the AI.
For the further advancement and improvement of AI usage, it is becoming increasingly important for companies to build an environment that connects to reliable and accurate data.
For Corporate Data Management, Choose uSonar
uSonar provides the "LBC" corporate database*, the largest in Japan, to companies through various cloud services. Corporate customer data, such as suppliers and sales leads, inevitably becomes obsolete over time and can be compromised by manual entry errors and inconsistent formatting. Companies that implement uSonar can build a data environment with comprehensive, accurate, and up-to-date information, as the high-precision data required for business is constantly supplied by our company.
For further details or to request materials, please use the inquiry form on our website.
*What is the "LBC" Corporate Database?
It is the largest corporate database in Japan, independently built by uSonar. (Largest in Japan: Based on our own research as of June 1, 2026)
We assign an 11-digit management code to approximately 12.5 million business locations across Japan. It maintains attribute information such as corporate numbers, industry types, sales scale, capital affiliation, and headquarters/branch relationships, allowing for the understanding of nationwide scale and affiliation structures. It covers a wide range of information, from offline data such as commercial registries to corporate information on the web, integrating and accumulating over 120 data points.
Contact for Inquiries:
uSonar Co., Ltd. Public Relations (Hiyama, Ehara)
Email:pr@usonar.co.jp
