.png)
2 days ago
2
70 million U.S. workers are about to face their biggest workplace transition due to AI agents, but their voices are too often missing. We address this gap by conducting a nationwide audit to understand what workers want AI agents to automate or augment, and how those desires align with the current technological capabilities. Data from 1,500 workers across 104 occupations leads to three main findings:
AI Agents are Reshaping the Workplace
You may not take interest in AI, but AI will take interest in you.AI is causing rapid, unexpected changes across the workforce. Studies estimate that around 80% of U.S. workers may see LLMs affect at least 10% of their tasks, with 19% facing potential disruption to over half of their responsibilities . Early-2025 LLM Usage data further indicates that AI tools are already in active use for at least 25% of tasks in 36% of occupations. To prepare for the future of work, in collaboration with economists from Stanford Digital Economy Lab, we propose a principled, survey-based and audio-enhanced auditing framework for mapping the risks and opportunities of AI agents across the full spectrum of U.S. occupations. The auditing framework takes a worker-centric approach by soliciting first-hand insights from domain workers actively performing corresponding tasks. Leveraging the U.S. Department of Labor’s O*NET database as the task source, we construct the AI Agent Worker Outlook & Readiness Knowledge Bank (WORKBank), the first database that captures AI agent capabilities and worker preferences. This database currently consists of responses from 1,500 workers across 104 occupations and annotations from 52 AI experts, covering 844 occupational tasks. It is designed to be easily extensible to more tasks and to reflect evolving technological capabilities and worker preferences.
Overview of the auditing framework and key insights.
Understand the Fear and Desire
1. Where do workers resist AI agent automation?
We analyzed worker transcripts using topic modeling, guided by the seed prompt: "The top most common fears that workers have about AI automation in their work." The three most prominent concerns identified are lack of trust (45%), fear of job replacement (23%), and the absence of human touch (16.3%). When breaking down WORKBank by sectors, in Arts, Designs, and Media, only 17.1% of tasks got positive ratings.
2. Which occupational tasks do workers desire AI agent automation?
For 46.1% of tasks, workers currently performing them express a positive attitude (rating their desire above 3 on a 5-point Likert scale) toward AI agent automation, even after explicitly considering concerns such as job loss and reduced enjoyment.
3. Why do workers want AI agent automation?
For pro-automation responses, we collected workers' motivations using both checkbox and free-form questions. The most cited motivation for pro-automation is “freeing up time for high-value work” (selected in 69.4% of cases). Other common reasons include task repetitiveness (46.6%), stressfulness (25.5%), and opportunities for quality improvement (46.6%).
4. Contrasting worker and AI expert perspectives delineate four task zones.
Our data helps classifies occupational tasks into four zones: 1. Automation “Green Light” Zone: Tasks with both high automation desire and high capability. These are prime candidates for AI agent deployment with the potential for broad productivity and societal gains. 2. Automation “Red Light” Zone: Tasks with high capability but low desire. Deployment here warrants caution, as it may face worker resistance or pose broader negative societal implications. 3. R&D Opportunity Zone: Tasks with high desire but currently low capability. These represent promising directions for AI research. 4. Low Priority Zone: Tasks with both low desire and low capability.
5. The desire-capability landscape reveals opportunities and mismatches.
We used Y Combinator (YC) companies as a proxy and mapped them to the tasks in WORKBank database. Unfortunately, the current YC investment does not skew towards Automation “Green Light” Zone and R&D Opportunity Zone. 41.0% of YC companies are mapped to Low Priority and Automation “Red Light” Zone; while many promising tasks within the “Green Light” Zone and Opportunity Zone remain under-addressed by current investments.
Opportunities for Human-Agent Collaboration
A distinctive aspect of our auditing framework is that it goes beyond the typical focus on automation. We also examine augmentation—where technology complements and enhances human capabilities. To provide a shared language for quantifying automation vs. augmentation, we introduce the Human Agency Scale (HAS), a five-level scale from H1 (no human involvement) to H5 (human involvement essential). This new scale complements the SAE L0-L5 automation levels by quantifying the degree of human involvement required for occupational task completion and quality rather than focus on an “AI-first” view.
Levels of Human Agency Scale (HAS).
6. Workers in many occupations prefer a balanced, collaborative partnership with AI.
We introduce the Human Agency Scale (H1–H5) to quantify the degree of human involvement required for completing occupational tasks and ensuring their quality. This new scale centers human agency and provides a shared language to capture the spectrum between automation and augmentation. Notably, Human Agency Scale H3 (Equal Partnership) emerges as the dominant worker-desired level in 47 out of 104 occupations analyzed.
7. Workers generally prefer higher levels of human agency, potentially foreshadowing frictions as AI capabilities advance.
Among 844 tasks, workers prefer higher levels of human agency than what experts deem technologically necessary on 47.5% of tasks. Notably, for 16.4% of tasks, the worker-preferred level is two levels higher than expert assessments.
8. Human Agency Scale reveals automation-vs-augmentation profile for each occupation.
Tasks within the same occupation can vary significantly in their desired levels of human agency. We suggest that AI agent development should account for varying levels of human agency to enable higher-quality and more responsible adoption.
Prepare for the Future
Not all types of work are equally impacted by AI. To understand where the future of work is headed and what skills will be most valuable, we further use the WORKBank database to analyze human skill shift. Using the O*NET database, we match each occupational task to the specific skills it relies on. For example, the task “Approve, reject, or coordinate the approval or rejection of lines of credit or commercial, real estate, or personal loans” (performed by financial managers) will be mapped to “making decisions and solving problems” and “guiding, directing, and motivating subordinates”. For each skill, we estimate two key values:
By comparing skill rankings based on these two dimensions, we uncover three emerging trends that could potentially shape the future of human work:
Comparing skill rankings by average wage and required human agency reveals a potential shift in valued human competencies---from information-processing skills to interpersonal skills.
