May 4, 2026
TOKYO – More than 30% of Japanese companies surveyed by The Yomiuri Shimbun and Teikoku Databank Ltd. reported using generative artificial intelligence in their business operations. Many said they used AI for tasks such as writing documents and gathering information.
When asked about concerns and challenges surrounding the use of generative AI, many of the surveyed firms cited accuracy of information and having a lack of specialized personnel or expertise.
The survey was conducted online from March 17 to 31, targeting 23,349 small, midsize and large firms nationwide. Of those contacted, 10,312 responded, resulting in a 44.2% response rate.
Of the 10,312 respondents, 34.6% said they had used generative AI “very much” or “somewhat,” while 14.2% said although they were not using it, they were considering its use in the future.
When asked what they used generative AI for, the largest group — at 45.1% — said writing, summarizing or proofreading texts, followed by information gathering at 21.8% and generating ideas when planning at 11%.
In contrast, 1.3% said delegating administrative tasks such as accounting and expense calculations and automating customer service by 0.5%.
When asked if they had experienced any negative impacts or issues regarding generative AI use, 67.7% of the companies who said they had used AI answered “no.” Multiple answers were allowed.
Among those who answered “yes,” the largest group — at 18.8% — said the gap had widened between employees who can effectively use generative AI and those who cannot in terms of capabilities and performance. While 4.5% said diversity among employees had decreased as they began submitting similar opinions or reports, 4% said employees had become dependent on generative AI for their work, leading to a decline in their motivation and skills, or capabilities.
When asked to make a list of up to three concerns or challenges regarding the use of generative AI, 50.4% cited the accuracy of information, followed by a lack of specialized personnel and know-how at 41.3% and scope of tasks suitable for the AI at 40%.