A recent study by the Stanford Institute for Human-Centered AI and the Digital Economy Lab explores the gap between what workers want from artificial intelligence (AI) and what it currently offers. Researchers surveyed 1,500 U.S. workers and interviewed 52 AI experts to understand where AI can benefit work and where it might be harmful.
The study found that workers desire automation mainly for repetitive tasks but want to maintain control over these tools. “As the workforce evolves, understanding and bridging the gap between worker expectations and the realities of AI capabilities will be crucial for organizations striving for successful integration,” said Diyi Yang, a Stanford assistant professor of computer science.
Trust in AI systems emerged as a major concern among respondents, with 45% doubting their accuracy and reliability. Additionally, many feared job loss or lacked confidence in human oversight. However, they welcomed automation that could free up time for more valuable work, reduce repetitiveness, and improve work quality.
Erik Brynjolfsson, director of the Stanford Digital Economy Lab, noted that “AI agents can play a supportive role in the workplace.” The research team categorized tasks into four zones based on desire for automation and technical capability: Green Light Zone (high desire and capability), Red Light Zone (low desire but high capability), R&D Opportunity Zone (high desire but low capability), and Low Priority Zone (low desire and capability).
Significant mismatches were identified between desired tasks for automation and current technical feasibility. This included writing creative content or preparing meeting agendas. “This map highlights a pressing need to intensify research efforts focused on tasks in the R&D Opportunity Zone,” Brynjolfsson emphasized.
The study also suggests that skills related to data analysis may decrease in value as AI capabilities grow, while skills requiring human interaction will become more important. Yijia Shao, a PhD student at Stanford leading this project, stated that bringing worker perspectives is critical for ethical adoption of AI systems.
Authors of this study include Yijia Shao, Humishka Zope, Yucheng Jiang, Jiaxin Pei, David Nguyen, Erik Brynjolfsson, and Diyi Yang.
For further details on this study visit the project website or see the paper on arXiv.
