"AI made my work easier. But somehow, I'm still tired." 62% feel AI fatigue, where did the time saved by efficiency go?

CHOIX Co., Ltd. conducted a "Survey on AI Utilization and Workload" targeting business professionals who utilize AI tools in their work, receiving responses from 200 individuals. [Survey Results Summary] * Approximately 90% feel that AI adoption has "made their work easier," confirming the widespread increase in efficiency. * Among those who found work easier, about 70% still experience AI fatigue, a paradox where efficiency and fatigue coexist. * AI fatigue is primarily caused by repetitive verification tasks and information overload, with increased decision-making burden being a contributing factor. * Those who found work easier are more creative in their use of AI, indicating that burden reduction is not naturally occurring. * Time saved by efficiency is allocated to "rest" or "other tasks," highlighting a gap between ideals and reality. [Detailed Survey Results] ◼️ Many people feel "AI makes work easier yet somehow tired," revealing a structure where efficiency and fatigue coexist. When asked about changes in workload due to AI tool utilization, 88.0%—nearly 90%—reported feeling a reduced burden, combining "much easier" and "somewhat easier" responses (n=200). Conversely, when asked about experiences of feeling "somehow tired," 62.0%—over 60%—reported experiencing AI fatigue, combining "often" and "sometimes" responses. Looking at the relationship between changes in workload and fatigue, even among those who answered "much easier," 40.2% "often" felt tired and 27.6% "sometimes" felt tired, meaning about 70% experienced fatigue. Furthermore, over 60% of those who answered "somewhat easier" also reported fatigue, revealing a structure where efficiency and fatigue coexist (n=200). ◼️ "Repetitive verification and judgment" is the most common cause at over 60%, with increased decision-making burden at the core of AI fatigue. When asked about the content of AI fatigue, "fatigue due to repetitive verification and judgment" was 64.3%, "fatigue due to information overload" was 51.4%, and "fatigue due to sustained tension" and "fatigue due to lack of tangible results" were both 25.7%. This suggests that as AI use increases, so do the demands for judgment, and the accumulation of these judgments leads to fatigue (n=179 / those who have experienced AI fatigue). Specific episodes related to AI fatigue included comments about "increased verification work," "trouble adjusting prompts," and "consistency with existing materials," revealing new burdens created behind the scenes of efficiency. * Rewriting prompts multiple times to get the desired output is stressful. Nuances that would be conveyed in a single word to a human are not understood, and the "effort to explain" exceeds the "effort to do it myself." * When verifying and correcting documents created by AI, I end up spending as much time as if I had created them myself. * When reviewing AI-generated deliverables that clearly look like they were made by AI, I worry about what kind of criticism I might receive from others, causing mental anxiety. * Although drafting contracts became quicker, working hours did not decrease, the overall workload increased, and I had to think more than before. * Discrepancies in wording and expressions between previously human-made documents and AI-made documents created more burden in reconciling consistency. ◼️ Approximately 70% make efforts to reduce AI fatigue. When asked about coping with AI fatigue, 32.0% said they were "making significant efforts," 39.5% "making some efforts," and 28.5% "making no efforts," indicating that about 70% are making some kind of effort (n=200). ◼️ Those who found work "easier" made more efforts, indicating that burden reduction is a "designed outcome." Looking at the relationship between changes in workload and efforts, among those "making significant efforts," 81.3% said work became "much easier," a conspicuously high figure. In contrast, only 22.8% of those "making no efforts" reported work being "much easier." This suggests that the benefits of AI are not gained merely by adoption; burden reduction depends on how one utilizes AI (n=200). Specific efforts included "batch processing of verification tasks," "using multiple tools selectively," and "clearly defining AI's assigned roles," showing an attitude of trying to reduce the burden by designing how AI is used. * Using multiple AIs selectively and choosing reliable tools. Switching to a different AI if one isn't working out. * Using multiple AI tools to derive several answers. * Clearly specifying roles and breaking down requests into steps to improve accuracy. * Avoiding excessive reliance on AI, creating documents based on templates oneself, and using AI as an aid for verification and checking tasks, thus consciously dividing roles. * Consolidating the review of output results rather than doing it piecemeal, reducing the burden. ◼️ Approximately 70% are conscious of how they spend their time, but completely... (The text is truncated here).