Labor Automation Forecasting Hub

Real-time forecasts from our global forecasting community on the future of the US workforce as AI advances.

Commentary last updated:
Metaculus forecast
Government baseline
Overall employment is projected to fall 0.8% by 2030 and fall 2.0% by 2035 relative to 2025 due to AI-driven displacement. This sharply contrasts with government baselines projecting +3.1% growth over the decade when accounting for aging-adjusted population trends.
Data fromon Jun 5, 2026

Activity Monitor

May 21, 2026
California Governor issues executive order to explore policies related to AI-driven workforce disruption - Office of Governor Gavin Newsom
May 15, 2026
Pope Leo XIV publishes an encyclical addressing human dignity in the era of artificial intelligence - The Vatican
May 14, 2026
Indeed Hiring Lab releases a report based on structural modeling documenting anticipated labor effects of AI - Indeed Hiring Lab
May 7, 2026
The Budget Lab research center at Yale published a study on the effects that AI tools are having on the labor market that found no clear effects as of yet - The Budget Lab at Yale

Jobs Monitor

AI is reshaping the job market, but not all fields are affected equally.

(Percentage change in employment)
Expected growth
Registered Nurses
+10.6%
Restaurant Servers
+8.2%
Law Enforcement
+7.1%
Construction Workers
+4.7%
Physicians
+3.7%
Engineers
+1.6%
By 2035, nurses, restaurant workers, and law enforcement are expected to see the highest growth, driven by hands-on needs that cannot be easily automated.
Expected decline
Designers
-0.8%
Janitors and Cleaners
-1.0%
Financial Specialists
-2.8%
Laborers and Movers
-3.5%
K-12 Teachers
-3.5%
General Managers
-6.6%
Services Sales Representatives
-7.9%
Software Developers
-8.1%
Lawyers and Law Clerks
-9.0%
By 2035, software, sales and law are expected to see sharp staff reductions as AI systems take over large portions of coding, outreach, and detailed research work, while laborers and financial specialists see a contraction as more advanced robotics roll out and AI automates accounting and administrative tasks.
Data fromon Jun 5, 2026
Bharat Chandar avatar
Bharat Chandar
Postdoctoral Researcher, Stanford Digital Economy Lab

Median overall employment forecast:
(2027: +1%) (2030: -0.5%) (2035: -4%)

In the very short run, I expect lags in employment impacts because of limitations of the technology and slow AI adoption. For this reason my 2027 estimate takes the trend line of employment and slightly undershoots it. However, power users may be as important to monitor as laggards because they may exert competitive pressure on markets that lead to faster adjustment.

In the longer run (5-10 years), I am extremely uncertain. I expect the technology will be much more advanced and integrated into peoples' lives. My primary uncertainty is the policy response and rate of worker adjustment if AI leads to rapid change. I don't know how this will resolve itself. There may be scenarios where AI does less than it is capable of because of new regulation. The BLS may also measure activities as work that look more like leisure than what many people do today.

Patrick Molgaard (draaglom) avatar
Patrick Molgaard (draaglom)
Pro Forecaster

The economic changes we see from AI will be faster than almost anything seen before. As a general trend, each technological wave is adopted faster than the previous (e.g. mobile phone penetration vs landlines) and the nature of AI should accelerate its adoption even relative to this trend.

Despite this, adoption and job displacement may still be surprisingly slow in some important senses. My stereotype of how this might look is that new AI-first competitor companies have been (or will be) created in many industries and these new entrants will take some time - a period of several years - to displace the old ones. As an intuition, "Photographic Process Workers and Processing Machine Operators" took 5 years between 2010 and 2015 to decline 50% - and this is a job whose associated technology was ~obsoleted.

Relatedly, I expect many job roles, even some seen as relatively "low education" or "at risk of automation" will have a surprisingly large long tail of tasks that take some time for AI systems to be good at. I'm also quite skeptical that a majority of the job losses attributed to AI so far (e.g. tech layoffs) are truly proximately caused by AI.

Hours, Pay, and Financial Well-Being

Forecasts currently show that despite a predicted decline in overall employment, median wages are expected to grow. The workweek is also expected to become about three hours shorter by 2035 among all workers, while productivity grows.

Lower income households are expected to see their government benefits outpace their basic needs, while higher income households are expected to see much stronger growth in resources relative to needs.

As AI automates more routine tasks, the question is whether the time freed up will translate into genuine leisure for workers or simply be filled with new demands, making the average workweek a key barometer of whether AI's productivity gains are actually shared with labor.

Forecasters note the concept of “dark leisure” may confound reported hours worked, as people may remain at work but do something else as they often have no incentive to transfer the time gains to their company.

The wellbeing measure below reflects a family’s available resources (after taxes, government benefits, medical expenses, childcare, and more), relative to a poverty threshold to meet minimum needs such as food, clothing, and shelter. These predictions show how the 20th, 50th, and 80th percentile families are expected to fare in the coming decade under the potential impact of AI, with higher numbers indicating better financial well-being.

With only 13% of workers using AI daily as of early 2026, the workplace is still in the early stages of an adoption curve that could fundamentally change how most Americans do their jobs within a decade. But forecasters note that some people may only think about AI as LLM chatbots and not realize how many tools they use in their daily work involve AI, especially as integrations increase across productivity tools.

Impact on the Next Generation of Workers

New college graduates are predicted to face difficult prospects in 2030 and beyond, as early-career tasks are more easily automated while experience and judgment remain harder to replace.

The unemployment rate for new graduates is expected to have more than doubled in 2035 compared to 2025.

The number of degrees awarded for STEM is expected to see only minor change due to the long gestation time, while overall 4-year degrees are expected to see a modest decline and humanities degrees are expected to see a more substantial decline by 2035. Trade schools and community colleges are expected to see significant growth in degrees and certificates awarded by 2035.

The rise of AI is threatening to accelerate an already-looming decline in 4-year college enrollment, as fewer high school graduates and shrinking job prospects for degree-holders could combine to reshape the future of higher education. At the same time, enrollment in community colleges and trade schools is expected to increase.

Yann Riviere (exmateriae) avatar
Yann Riviere (exmateriae)
Pro Forecaster
One thing that is pretty important is how fast the capabilities takeoff will be. If you're expecting a very fast takeoff, then many of the 22-27 may find themselves out of any opportunity very soon. If things continue to ramp up like they're currently doing, young people may be able to pivot to other areas soon enough to increase their chances of finding a job.

Changing Economy

As AI capabilities continue to grow, the increase in productivity and automation of jobs are likely to lead to significant changes in the overall economy.

Companies will be able to make money with fewer employees, meaning both higher unemployment and less revenue sharing with the labor force.

While the forecasted changes may seem relatively modest, at the macroeconomic level these represent notable shifts that will likely have ripple effects throughout society.

AI is expected to enable companies to generate more revenue with far fewer employees, and over the next decade a growing share of Fortune 500 giants could operate with workforces once associated with small businesses rather than corporate behemoths.

Ľuboš Saloky (lubossaloky) avatar
Ľuboš Saloky (lubossaloky)
Pro Forecaster

Even during periods when total unemployment rates spike significantly, the rate of long-term unemployment relative to the labor force stays relatively low. People do exit the unemployment statistics without finding employment. When workers become discouraged and stop looking for employment, they leave the labor force. Also when someone transitions from being unemployed to returning to school, retiring early, or focusing on family care, they disappear from unemployment statistics.

When discouraged workers fall off the unemployment rolls, the unemployment rate looks artificially lower. I’m forecasting a −2% change in overall employment by 2030 and −11% by 2035.

However, I don’t expect these declines to be fully reflected in the long-term unemployment rate.

If AI allows corporations to generate ever-greater output without proportionally growing their workforce, workers could claim a shrinking slice of the economic pie, marking a redistribution of income from labor to capital. Even a few percentage points marks a major shift in the context of historical trends.

Comparison to Existing Research

A number of recent research publications have identified occupations, tasks, and industries that are more exposed or vulnerable to automation, while other work has examined recent employment patterns for signs of AI’s labor market effects. Recent work from Stanford University argues that AI has already had an impact on early career work, while other sources such as research from the Budget Lab at Yale do not yet see strong signals. Exposure and vulnerability ratings typically are not intended to be predictive of the future, but instead are correlational measures of current AI usage and task patterns.

The forecasts Metaculus is presenting in the Hub fill a gap in our current understanding, directly providing wisdom of the crowd powered predictions on employment outcomes when taking into account the impact of AI. In many cases, the forecasts align with what the exposure and vulnerability literature would indicate, with some key differences.

In the AI exposure literature and research, teachers stand out as having high exposure ratings, but are predicted to see mild decline as forecasters expect human presence will be desired in classrooms by schools and parents, even if schools do increasingly adopt AI-powered educational tools. Conversely, warehouse workers are rated as low exposure due to the high physical nature of the work, but forecasters anticipate that robotic capabilities will begin to displace more of these roles by 2035. Forecasters do expect that the high exposure and vulnerability of lawyers, sales representatives, financial specialists, and software developers will translate to significant employment reductions in these fields over the next decade. These forecasts provide important context to our understanding of workforce prospects by quantifying the predicted impact of AI on employment levels.

Adonis da Silva (Adonis) avatar
Adonis da Silva (Adonis)
Pro Forecaster
Even if some professions become completely obsolete, employment levels could continue along the same trend if people shift to other careers. Demand is likely to increase in jobs that rely on human interaction or in which products increase in value if they're human-made or scarce, or, temporarily, in some that require manual labor.
Note: each literature column has its own color range, where the reddest is the highest AI exposure number in the column and the greenest is the lowest AI exposure number in that column. The Metaculus forecast columns for 2030 and 2035 share a color range; the largest predicted decline across 2030 and 2035 is the reddest, while the largest predicted growth across 2030 and 2035 is the greenest.
Employment declineEmployment growth
Higher AI exposureLower AI exposure
  1. Occupational exposure figures from Felten et al. (2023), estimating how much the typical tasks and abilities in each occupation overlap with what generative AI systems are good at. Data was based on 2010 SOC codes, which Metaculus has crosswalked to the 2018 SOC. The paper presents both language modeling and image generation exposure scores, from which only language modeling scores are used here.
  2. A vulnerability score calculated from Manning and Aguirre (2026) from measures of AI exposure and adaptive capacity (a measure of a worker’s ability to navigate job transitions if displaced). These scores were combined into a vulnerability score using the same approach as used in Figure 1 of the paper.
  3. Occupational exposure measured by Anthropic using usage data from their AI system, Claude, as reported in the data for the Anthropic Economic Index. More details about Anthropic’s findings are reported in Massenkoff and McCrory (2026).
See the underlying data and how we adapted the figures from these sources here.

State-Level View

To complement the national forecasts, we also look at the state of Washington to see whether short-term expectations at the state level track the broader national pattern. Washington is especially useful as a test case because of its concentration in dynamic industries like technology and aerospace, which could potentially see more dynamic short term changes.

The healthcare sector (employing 13% of Washington residents) is forecasted to grow through 2027, largely unaffected by AI in this timeframe. Aerospace (employing 2%) is expected to see minor growth in the short-term, while technology (employing 10%) is expected to see marginal growth, largely consistent with historical trends. These forecasts are short-term, leading to minimal predicted change.

Ľuboš Saloky (lubossaloky) avatar
Ľuboš Saloky (lubossaloky)
Pro Forecaster
AI adoption is expected to have only a small impact on employment levels before 2027. Washington's heavy concentration in technology amplifies this risk. Microsoft recently laid off 3,200 employees, while other tech companies such as Google and Meta have also reduced their workforce. Other industries, like manufacturing and aerospace, face mixed prospects. However, while the U.S. population grew by 0.5% from July 2024 to July 2025, Washington's rose by 0.9%. Even modest employment growth of roughly 0.7% annually, caused by the state's population expansion, should offset the uptick in unemployment. The growing population will generate sustained demand for workers in multiple industries, for example retail, healthcare and construction.

Methodology

The forecasts presented on this page were designed to address key uncertainties about the future impact of artificial intelligence on labor in the United States. They are produced by aggregating many individual forecasts into a prediction that research has shown to be more accurate on average than individuals typically produce. The sections below provide more details about how the information above was produced.

Reach out to Metaculus

Have thoughts, questions, or spotted something unclear? Your input helps us improve this dashboard.

View Services

Share the dashboard

Think this could help others understand how AI affects jobs? Share it with your network.

Share on X

Subscribe for updates

Sign up to get notified when we publish new forecasts, insights, and improvements.