What impact is artificial intelligence having on the U.S. labor market and the nation’s economy?
As new and groundbreaking technologies emerge, they can function as a blank canvas onto which people project both aspirations and anxieties. Artificial intelligence—despite its quick uptake in broader society and its rapid evolution so far—is still loosely defined in the public imagination and crosscutting in ways that can make it difficult to discern informed insight from speculation.
Yet AI is already impacting the U.S. economy, as well as current and prospective workers in myriad ways, meriting policymakers’ attention. While researchers are rapidly expanding the evidence base needed to better understand and govern AI, much of the field is still emerging.
This column looks broadly at artificial intelligence and its current uses in the U.S. economy, as well as potential areas of expansion in the future. It then turns to how policymakers can consider developing guardrails to protect the American public, particularly workers, from possible harms caused by AI or resulting from its development. This document is intended as a compass, rather than a definitive guide, to help policymakers navigate toward reliable, evidence-based sources, help frame key questions for policy development, and critically evaluate claims relating to AI and the economy.
What is artificial intelligence?
Artificial intelligence refers to a wide range of related technologies, primarily computer systems that can perform tasks without significant human oversight. AI is an umbrella term that includes common applications that have been supporting worker productivity for years, such as those that provide grammar and spelling support, bibliography generation, and project management. It also includes more sophisticated technologies, including large language models, such as ChatGPT, and software systems known as AI agents that can take over more time-intensive tasks.
For the purposes of this column, the 2019 National Defense Authorization Act provides a useful AI shorthand definition: “an artificial system designed to think or act like a human, including cognitive architectures and neural networks.” A more recent definition can be found in Executive Order 14110, issued in 2023, which defines artificial intelligence as a machine-based system that can, for a given set of machine- and human-based objectives, make predictions, recommendations, or decisions influencing real or virtual environments in an automated manner.
For different ways to think about AI, see the Equitable Growth essay by Sorelle Friedler of Haverford College and Harvard University’s Marc Aidinoff.
How is AI affecting the U.S. economy?
OpenAI Inc., the firm behind ChatGPT, recently claimed that AI will “drive unprecedented economic growth” in the U.S. economy in the long run. But it is essential to consider that a range of factors beyond technology—including public and private investments, productivity, and the makeup of the labor force—drive economic growth. Such factors intersect with AI technologies in differing ways and varying timescales.
How much investment in AI is occurring?
Proposals for how to measure the contributions of AI technology production to economic growth are emerging. By 2024, corporate AI investments were more than $250 billion. These investments in building out AI infrastructure are driving national Gross Domestic Product growth and raise concerns that the investments may not be able to provide an appropriate return—or worse, become a speculative bubble. During the first half of 2025, half of the country’s reported GDP growth, 1.2 percent, was reportedly due to the investments related to AI infrastructure. But the increasingly circular web of investments between the largest private firms investing in AI should be closely watched.
Few firms have reported any returns on those investments. A recent paper published by NANDA at MIT finds that of the tens of billions of dollars that private enterprises have invested, 95 percent of them reported no measurable impact on their profits and losses, primarily due to a lack of a productive learning feedback loop to improve pilot programs for workers’ actual workflows. Although workplaces are quickly adopting these new technologies, firms are not seeing the payoff to their bottom lines yet. AI technologies will need to become more productivity-enhancing before they have a meaningful effect on worker productivity and GDP, with current estimates of the impact being limited.
How is AI being deployed, and what impact is it having on workers?
AI firms’ own research finds that their large language models are primarily being utilized for software development and writing tasks to support existing workers through the augmentation of their jobs. But there also has been a recent rise in AI agents, or active assistants that can initiate tasks, adapt a plan to new information, and operate across applications. In some cases, augmentation can help reduce workers’ exposure to dangerous tasks, while in other cases it may make their jobs more challenging.
The introduction of self-check-out machines in lieu of traditional cashiers, for example, has reduced the number of guest service workers needed during retail transactions and created a new need for human interaction when the machine does not work as designed. This is not the result of AI technologies, but it shows how limiting worker interactions with customers to almost exclusively when an issue arises results in workers more likely to deal with frustrated customers.
In the spring of 2024, a study of U.S. Census Bureau data reported that only 5.4 percent of firms were using AI technologies, up from 3.7 percent the year prior. At the same time, recent research by Alexander Bick at the Federal Reserve Bank of St Louis, Adam Blandin at Vanderbilt University, and David Deming at the Harvard Kennedy School suggests that workplace adoption of generative AI—a specific AI category of “deep-learning models” that create content from existing data—has been swifter than the adoption of the personal computer. They report that almost 40 percent of the U.S. population has used generative AI at least once in the prior week, and 23 percent of those who were employed reported using AI at least once at work in the previous week.
Still, the three co-authors suggest that AI technology applications will take some time to refine in order to support widespread and impactful increases in productivity. Their analysis suggests that the time savings AI currently provides amounts to only 1.4 percent of work hours. And researchers at the Massachusetts Institute of Technology’s Shaping the Future of Work initiative estimate that the number of workers who may lose their positions to AI-driven automation to be as low as 1 percent to 2 percent of the workforce over the next 20 years. Yet one of the purveyors of the technology itself—AI firm Anthropic’s CEO Dario Amodei, who has a sizeable financial stake in selling this version of the future—claims it could drive unemployment to record highs in the immediate term.
Indeed, some employers have already implemented aspects of AI technologies into their employees’ workdays without workers’ knowledge or consent in an attempt to increase productivity, producing negative and unwanted effects such as limited productivity gains and the diminished health and safety of workers. Survey research by Alexander Hertel-Fernandez at Columbia University suggests that employers’ adoption of AI-enabled automated management and surveillance is already fairly widespread in the U.S. economy: An estimated 70 percent of workers across sectors and occupations in the economy may already be subject to some form of electronic monitoring by their employers, leading to negative impacts on workers’ health and well-being.
Large language models also can produce hallucinations, or inaccurate content that can be made up or false. Such hallucinations can contribute to a proliferation of misinformation or present direct risks when providing incorrect and dangerous professional advice. Sometimes AI-generated work is poorly prepared, referred to as “workslop,” thus limiting, or even diminishing, the productivity gains the technology claims to provide.
The introduction of AI into management software is already having profound effects on some worker’s abilities to earn fair wages. Research into AI labor-management vendors in various sectors, from health care to logistics, finds that employers are using the technology to set compensation structures and calculate individuals’ wages, leading to an erosion of fairness and transparency in setting and calculating workers’ pay, as detailed by law professor Veena Dubal at the University of California, Irvine and her co-author Wilneida Negrón, in “How artificial intelligence uncouples hard work from fair wages through ‘surveillance pay’ practices—and how to fix it.
How is it impacting the U.S. labor market?
JP Morgan Chase & Co. CEO Jamie Dimon has already blamed the adoption of artificial intelligence for job reductions and reduced demand for workers, but the research is mixed. Stanford University’s Digital Economy Lab finds that employment for young workers meaningfully decreased in occupations that are particularly exposed to AI, while other work suggests that recent graduates’ unemployment rates were already climbing prior to the introduction of large language models in workplaces. New analysis by the Yale Budget Lab of Current Population Survey data suggests that the introduction of large language models has not yet caused a decrease in the need for cognitive labor.
Whatever the case, the perceived downturn in U.S. economic activity recently could be due to a mix of tariffs, inflation, immigration crackdowns, reduced rates of job creation, and low consumer confidence, rather than any AI-induced effects on the U.S. labor market. Making these assessments more difficult is the fact that the federal government’s statistical agencies are not currently equipped to track AI’s impact on the workforce, while the recent disruptions at the U.S. Department of Labor’s Bureau of Labor Statistics mean that recent GDP data are currently unavailable.
How do we anticipate AI will affect the U.S. labor market in the future?
The widescale adoption and integration of AI into the U.S. economy is still in its early stages, and it is not clear whether extensive automation will occur since AI is both augmenting and automating tasks within jobs. It also is not obvious how slowly or quickly AI adoption in the workplace will happen, though evidence from NANDA at MIT suggests firms are striving to swiftly integrate the technology into their workflows.
The pace by which AI is adopted into the workplace will be shaped by innovation, market and regulatory forces, and human preferences. More widespread use also will require that the professional and personal adoption of AI technologies continues to increase, the public and private sectors operate as rational actors that seek to maximize their benefits and minimize their costs, and energy production keeps pace with AI’s incredible need for power.
Some jobs are more susceptible to exposure to artificial intelligence technologies than others. Researchers can provide us with some insight into how susceptible an occupation might be based on whether AI technologies can perform core aspects of that job. Consider a primary or secondary school teacher: AI technologies may currently be able to handle some tasks, such as drafting lesson plans or building a curriculum, but not other tasks, including real-time classroom management or supporting children to develop their emotional-social competencies.
Jobs with fewer tasks are more at risk of being automated by AI technologies independent of human supervision compared to jobs that require human interaction. If AI can perform the core functions of an occupation, then it is much more susceptible to automation and replacement, while if AI can only perform supplemental tasks, then workers enjoy more stability. In a recent Equitable Growth report, “Adoption of generative AI will have different effects across jobs in the U.S. logistics workforce,” the four co-authors explore this distinction in more detail.
Exposure to AI also doesn’t seem to be equally distributed. A new working paper and accompanying column by Equitable Growth’s Chris Bangert-Drowns and Chiara Chanoi suggests that workers with higher levels of education in high-paying positions are particularly exposed to AI, and women tend to work in more exposed occupations. This research builds on recent work by the Pew Research Center and Anthropic.
AI could impact workers in ways other than via automation or augmentation. AI could produce a workplace where jobs are broken up by task and translated into roles closer to those considered part of the gig economy today. This fragmentation of jobs could happen by breaking apart roles into discrete tasks and using AI technologies for some aspects of the position and human workers for others.
During the recent Writers Guild of America’s strike, for example, there was a concern that large Hollywood studios would replace human writers and utilize AI to write first drafts of scripts, with the human writers coming back as editors, shortening their time of employment and lowering their pay. Ultimately, it was the strength of the Writers Guild of America that staved off the worst potential impacts and job degradation of AI on their union members.
This is not to say that AI will not disrupt the U.S. workforce. AI-driven innovations could lead to sizable wage declines for concentrated groups of workers in certain occupations most at risk, such as office and administrative support, financial operations, and media occupations.
What should policymakers consider?
Although there are many unknowns—and new innovations such as Open AI’s ChatGPT-fueled Atlas web browser are seemingly constantly being unveiled—the information available at this stage of AI’s adoption and deployment suggests that there is a need for federal leadership to provide some rules of the road. Policymakers at least need to consider how to regulate the use of AI to protect workers, consumers, and data privacy, while also promoting innovation and market competition and minimizing public risks. Attempts to prevent thoughtful governance are short-sighted and put the economy and public safety at risk.
Computer scientists Arvind Narayan and Sayash Kapoor at Princeton University argue that policymakers should view “AI as normal technology,” rejecting the vision that AI is itself an agent in determining the future. Policymakers should heed their recommendation and view AI as a tool that can be controlled. Below is an initial set of policy considerations that Equitable Growth will continue to build on in the coming months.
What guardrails are in place to guide the integration of AI in the economy?
After the Trump administration rescinded and replaced the Biden administration’s executive order on artificial intelligence and the other related executive orders on AI, there are now few guardrails in place at the federal level. Critically, some researchers have already detailed how the introduction of AI-powered algorithmic wage discrimination practices have been violating existing fair wage laws, as well as health and safety protections, without facing federal or state enforcement.
In the absence of strong federal leadership to regulate AI as it proliferates in our everyday lives, there has been a flurry of activity at the state level across a range of topics and issues that intersects with the introduction of AI to society. These bills and laws illustrate the need for comprehensive congressional action to provide guidelines rather than deregulation or a full moratorium on state-level innovations that could yield important insights for advancing federal efforts.
Arguments in favor of commercial deregulation that frame AI innovation as a race with China or other U.S. competitors under the guise of national security and as a zero-sum game should be viewed with skepticism. Racing to be first in AI innovation with no guardrails could produce dangerous results by encouraging all parties to cut corners, be more secretive, and risk consumer safety in the name of speed. Nobel laureate Daron Acemoglu at MIT (and a member of the Washington Center for Equitable Growth’s Steering Committee) has detailed how more deregulation does not spur innovation, but rather how rules and regulated markets attract investment by mitigating risk and supporting optimal decisions.
Indeed, strong institutions, predictability, and programs that help spur research and development that would be cost-prohibitive without public-private partnership funding for university-based research and development (included in the 2022 CHIPS and Science Act and the Inflation Reduction Act) are essential for innovation. Many of these programs, however, are currently at risk of losing or have already lost vital federal funding.
How do we address AI market concentration?
Left unchecked, economic activity surrounding the rollout of AI technologies could dampen economic growth by reducing employment dynamism, in addition to other negative externalities, such as the sizeable environmental costs and a greater concentration of wealth at the top. A healthy economy needs fair and competitive markets, achievable through active antitrust enforcement.
Given the small number of large private firms dominating the market for AI products, such as large language models, there are fears that the technology stack that supports AI is increasingly concentrated, as described by University of California, Berkeley law professor Tejas Narechania in “Understanding market concentration in the AI supply chain.” This market dynamic has led researchers and lawyers to pursue frameworks for taxing revenue streams produced by data and AI services to democratize some of the positive economic impacts that AI could produce. Mona Sloane at the University of Virginia and Ekkehard Ernst, chief economist at the International Labour Organisation, explore this in “Tackling AI, taxation, and the fair distribution of AI’s benefits.”
If the U.S. labor market experiences large disruptions, what will happen to workers?
There could be a large disruption to the U.S. workforce, in which swaths of workers become unemployed, for many reasons, among them an economic recession or public health crisis. Given the currently weakening U.S. labor market, future layoffs, which may or may not be the direct result of the adoption of AI, could be attributed to the technologies all the same because it is difficult to distinguish between larger market forces, corporate consolidation, and AI adoption in pinpointing labor market effects.
Even without large-scale job losses, the current U.S. safety net is inadequate to help workers in the immediate term since Unemployment Insurance only replaces a portion, typically less than 50 percent, of workers’ average weekly income for up to 26 weeks, though many states are less generous. Of the nearly 4 million unemployed individuals who were actively seeking employment in August 2025, more than a quarter had been unemployed for more than 27 weeks, according to the U.S. Bureau of Labor Statistics.
This number of unemployed workers could rise sharply during future periods of prolonged financial and economic instability. The economic changes enacted through the Republican-led HR1, including massive federal funding cuts and more stringent work requirements for some income support programs, such as Medicaid and the Supplemental Nutrition Assistance Program, will exacerbate holes in the safety net and shrink its efficacy.
The changes to these and other social infrastructure programs could not come at a worse time for the adoption of AI in U.S. workplaces. There will be short-, medium-, and long-term demand for these programs, whether AI is actually creating problems in the economy or just being blamed for them. Economic crises require government intervention, including emergency financial support for impacted workers and structures that optimize the nation’s human capital in the long run, such as meaningful pathways to retrain and reenter the workforce.
Some policymakers are aware of this potential need. U.S. Sen. Mark Kelly (D-AZ) recently proposed establishing the AI Horizon Fund to provide support for initiatives that train and reskill the U.S. population and prepare for changes that AI-driven innovation might bring to the U.S. economy.
How can we protect workers’ rights?
One way to ensure that workers benefit from these advancements in technology is to include their voices when deciding upon businesswide changes that impact their jobs. In Germany, labor unions have helped produce optimal outcomes for workers and businesses alike, as detailed by Virginia Doellgast of Cornell University and Nell Geiser, research director at the Communications Workers of America, in their “Boosting U.S. worker power and voice in the AI-enabled workplace.”
Equitable Growth recently announced funding for additional research to further investigate the potential benefits of including workers’ voices in AI adoption in their workplaces, including a case study of the Labor Management Partnership at Kaiser Permanente and the role of worker voice in management strategy and job quality.
Conclusion
There is a lot that policymakers, academics, workers, and employers alike still do not know about how AI is directly and indirectly impacting the U.S. economy now and into the future. That is why Equitable Growth recently announced investments in 12 scholars to help generate actionable insights that policymakers can use as they navigate the era of AI innovation.
Innovation and economic growth do not have to be at direct odds with workers’ rights and economic equity. Artificial intelligence can complement workers’ job skills rather than replace them by centering workers in production processes when AI is intended only to augment their occupations. And researchers and technologists can work to ensure the ethical production and adoption of AI applications, as well as increase AI-relevant skills and literacy, so that productivity gains and benefits of artificial intelligence are more broadly shared.
Additional resources available to readers on AI, tech, and the economy.
Did you find this content informative and engaging?
Get updates and stay in tune with U.S. economic inequality and growth!
Stay updated on our latest research