The Future of Work with AI Agents: What Workers Actually Want
A groundbreaking study reveals the gap between AI capabilities and worker desires across 844 occupational tasks
As AI agents become increasingly sophisticated, one critical question remains: What do workers actually want from AI automation? A comprehensive new study from Stanford University provides the first large-scale answer, surveying 1,500 workers across 104 occupations to understand the evolving landscape of human-AI collaboration in the workplace.
Beyond the "Will AI Take My Job?" Question
Rather than asking whether AI can automate jobs, researchers took a different approach: they asked workers what they want AI to automate. The results challenge common assumptions about AI adoption and reveal a more nuanced picture of the future workplace.
Key Finding: Workers express positive attitudes toward AI automation for nearly half (46.1%) of their tasks—but not for the reasons you might expect.
What Workers Really Want from AI
The study introduced an innovative framework called the Human Agency Scale (HAS), which moves beyond simple automation to measure five levels of human involvement:
- H1: AI handles everything independently
- H2: AI needs minimal human input
- H3: Equal human-AI partnership (most popular choice)
- H4: AI requires significant human guidance
- H5: Continuous human involvement essential
Surprisingly, 45.2% of occupations favor H3—equal partnership—suggesting workers don't want to be replaced, but rather supported.
The Top Motivation: Time for Higher-Value Work
When workers expressed desire for automation, their primary reason wasn't laziness or job avoidance. 69.4% said they wanted AI to "free up time for high-value work." Other common motivations included:
- Task repetitiveness (46.6%)
- Opportunities for quality improvement (46.6%)
- Reducing stress (25.5%)
The Desire-Capability Landscape: Four Critical Zones
By mapping worker desires against expert assessments of AI capabilities, researchers identified four distinct zones:
🟢 Automation "Green Light" Zone
High worker desire + High AI capability = Prime automation candidates
🔴 Automation "Red Light" Zone
Low worker desire + High AI capability = Proceed with caution
🔬 R&D Opportunity Zone
High worker desire + Low current capability = Research priorities
⚪ Low Priority Zone
Low desire + Low capability = Not urgent for development
Critical Insight: 41% of Y Combinator AI startups are focusing on the wrong zones—building capabilities workers don't want rather than addressing high-demand areas.
Industry Patterns: Where Workers Resist AI
The study revealed stark differences across sectors:
- Arts, Design, and Media: Only 17.1% of tasks received positive automation ratings
- Software Development: Much higher acceptance rates
- Data Processing: Strong desire for automation of repetitive tasks
What Workers Fear Most
Among participants expressing concerns about AI, the top fears were:
- Lack of trust in AI accuracy/reliability (45%)
- Fear of job replacement (23%)
- Loss of human qualities in work (16.3%)
Creative professionals were particularly vocal about maintaining the "human touch" in their work.
The Skills Revolution: From Information to Interpersonal
Perhaps most significantly, the study suggests a fundamental shift in valuable workplace skills. As AI handles more information processing tasks, human skills are becoming more important in areas requiring:
- Interpersonal communication
- Organizational abilities
- Complex decision-making
- Creative problem-solving
Traditional high-wage skills like data analysis are becoming less emphasized, while human-centered capabilities gain importance.
Real Worker Voices
The researchers collected audio interviews revealing how workers envision AI collaboration:
"I want it to be used for seamlessly maximizing workflow... making things less repetitive and tedious. No content creation." — Art Director
"[I envision AI] as an assistant who is doing research for me. However, I review every answer because we cannot rely on its accuracy." — Research Professional
Implications for the Future
This research suggests three key takeaways for both workers and organizations:
- AI adoption will be more collaborative than replacement-focused
- Worker preferences should drive AI development priorities
- Success requires aligning technological capabilities with human desires
The study's WORKBank database provides a roadmap for responsible AI development—one that puts worker needs at the center rather than treating them as an afterthought.
The Bottom Line
The future of work isn't about AI replacing humans—it's about finding the right balance of human agency and AI capability for each specific task. Workers are ready to embrace AI, but on their own terms, focusing on augmentation rather than automation.
As AI agents continue to evolve, this research provides crucial guidance for building systems that workers actually want to use, creating a future where technology enhances rather than threatens human potential in the workplace.
The complete WORKBank database covers 844 tasks across 104 occupations and represents the first comprehensive audit of worker preferences for AI automation. The research was conducted between January and May 2025 by Stanford University researchers.
Reference
Original Paper: Shao, Y., Zope, H., Jiang, Y., Pei, J., Nguyen, D., Brynjolfsson, E., & Yang, D. (2025). "Future of Work with AI Agents: Auditing Automation and Augmentation Potential across the U.S. Workforce." arXiv preprint arXiv:2506.06576v2.
