Despite the widespread deployment of artificial intelligence by companies to improve efficiency, a new study has found a new type of workplace fatigue called 'AI brain drain' hidden behind it. An analysis report released by the Boston Consulting Group in conjunction with the University of California and the Harvard Business Review, based on a survey of nearly 1,500 full-time U.S. employees, found that frequent switching between AI tools, writing prompts, and monitoring automated processes actually increased employees' cognitive burden.
Many respondents described experiencing symptoms such as slow thinking, inattention, headaches, and even mental 'hangovers' after high-intensity AI use. Employees in marketing and human resources roles reported these symptoms particularly significantly, suggesting that when work requires continuous interaction with multiple AI systems, tools originally intended to reduce burden become new sources of tasks.
The study points out that modern enterprises often deploy multi-agent systems, and employees need to switch frequently between different platforms, data sources, and prompts. This 'managing AI' operation itself has evolved into a core energy-consuming link in the work, rather than a real efficiency improvement. If there is a lack of systematic governance, the auxiliary advantages of AI may be offset by cognitive overload, leading to increased decision-making errors, decreased thinking speed, and significantly reduced job satisfaction.
Data further shows that employees experiencing AI brain drain are 33% more likely to experience decision fatigue and about 40% more likely to voluntarily leave their jobs. Researchers estimate that if this type of efficiency loss and talent loss spreads on a large scale in large enterprises, it could cause millions of dollars in hidden costs each year. This phenomenon not only exists in traditional industries, but also quietly emerges in crypto and fintech teams - even in environments that pursue rapid iteration, the pressure to maintain system stability and security is still increasing the cognitive load.