# Future of Work with AI Agents: Auditing Automation and Augmentation Potential across the U.S. Workforce

**Authors:** Yijia Shao, Humishka Zope, Yucheng Jiang, Jiaxin Pei, David Nguyen, Erik Brynjolfsson, Diyi Yang  
**arXiv:** [2506.06576](https://arxiv.org/abs/2506.06576)  
**Date:** 2025-06-06  
**Summary page:** https://www.summarizepaper.com/en/arxiv-id/2506.06576v1/

## Key points

- Compound AI systems, or AI agents, are causing a significant shift in how work is performed
- Concerns about job displacement and reduced human agency due to automation
- Lack of systematic understanding of the changing environment
- Researchers led by Yijia Shao, Humishka Zope, Yucheng Jiang, Jiaxin Pei, David Nguyen, Erik Brynjolfsson, and Diyi Yang have developed a framework to assess worker preferences for automating or augmenting tasks
- Framework includes an audio-enhanced mini-interview method and a standardized language for quantifying human involvement in automated tasks
- The team has created the Automation "Red Light" Zone, R&D Opportunity Zone, and Low Priority Zone to classify tasks based on worker desires and technological capabilities
- Results reveal diverse profiles reflecting varying expectations regarding human involvement in different tasks
- Integration with AI agents may reshape core human competencies towards interpersonal skills as technology advances
- Importance of aligning AI agent development with human desires and preparing workers for evolving workplace dynamics

## Summary

In the rapidly evolving landscape of the labor market, the emergence of compound AI systems, also known as AI agents, is causing a significant shift in how work is performed. This transformation raises concerns about job displacement and reduced human agency due to an overreliance on automation. However, despite these challenges, there is a lack of systematic understanding of this changing environment. To address this gap, a team of researchers led by Yijia Shao, Humishka Zope, Yucheng Jiang, Jiaxin Pei, David Nguyen, Erik Brynjolfsson and Diyi Yang have developed a novel . This framework aims to assess which tasks within various occupations workers desire  to automate or augment and how these preferences align with current technological capabilities. Central to this framework is an audio-enhanced mini-interview method that captures nuanced worker desires. Additionally, the researchers introduce the  as a standardized language for quantifying the preferred level of human involvement in automated tasks. Through the implementation of this framework, the team has created the , leveraging data from the U.S. Department of Labor's O*NET database. This comprehensive database incorporates preferences from 1,500 domain workers and capability assessments from AI experts across more than 844 tasks spanning 104 different occupations. By jointly considering both worker desires and technological capabilities,< kd >Automation "Red Light" Zone,</ kd > < kd >R&D Opportunity Zone,</ kd >and< kd >Low Priority Zone.</ kd > These classifications highlight critical mismatches as well as opportunities for further development of AI agents. Moving beyond a simplistic binary view of automation versus non-automation decisions, the results obtained reveal diverse  These profiles reflect varying expectations regarding human involvement in different tasks. Furthermore, this study provides early insights into how integration with AI agents may reshape core human competencies within the workforce. There is a noticeable shift from information-focused skills towards interpersonal skills as technology continues to advance. Overall, these findings emphasize the importance of aligning AI agent development with human desires and preparing workers for the evolving dynamics within workplaces. The research conducted by this team sheds light on crucial considerations for navigating the future of work alongside AI technologies effectively.