Skip to main content

AI and the Workforce: What We’ve Learned and Areas for Exploration


Artificial intelligence is a pivotal technology, and its impact on the American workforce is a major point of interest for policymakers and the public. To this end, the Bipartisan Policy Center is conducting preliminary research and convening stakeholders and experts from business, labor, civil society, and academia to explore the early evidence and areas of debate around the impacts of AI adoption on the workforce.[1] This resource outlines lessons learned so far and identifies opportunities for additional research to better understand how AI may shape the future of work.


The impact of AI on the workforce will heavily depend on its adoption by companies and other employers. Though interest in AI has recently surged as developments in computing power and memory have dramatically augmented the capabilities of AI-powered tools, some companies have utilized AI systems in their workflows for decades. An early example of AI adoption occurred in the credit card industry when Visa began using AI to detect credit card fraud in 1993.

The productivity-boosting potential of AI is a major reason many businesses are exploring applications for AI in the workplace. Today, many companies are testing and deploying AI systems to improve their processes and decision-making capability while trying to figure out what works and what does not. For instance, human resource professionals use AI to help with recruiting and hiring. Call centers use AI to help with customer support. Software companies use AI to help them write computer code. However, the full extent of AI adoption in these industries remains unclear.

A Census Bureau survey conducted in late 2023 found that 3.9% of American businesses use AI to create goods and services, and only 6.5% plan to adopt AI in the following six months. The survey also reveals much higher adoption rates in industries like information (13.8%) and professional, technical, and scientific services (9.1%) relative to accommodation and food services (1.2%) and construction (1.2%).

Census Ai Graph

It is important to note that this survey data may not paint a full picture, in part because those businesses that have already adopted AI likely account for a much larger share of employment. An academic we spoke with highlighted how AI adoption is more common among larger firms and small startups that are “AI natives” based on her research, but she also added that AI adoption in production has been “low, slow, and extremely uneven.”  Additionally, many businesses eager to leverage AI—particularly small businesses—face barriers to adopting AI systems. A recent BPC survey of small business owners found that costs, regulatory uncertainty, data privacy, and lack of AI literacy among both owners and employees were some of the greatest barriers to AI adoption.

As more businesses seek to integrate AI applications, the potential workforce impacts of widespread AI adoption are top-of-mind for employers and employees alike. For instance, recent labor disputes in the creative industry focused in large part on AI’s impact on workers. A May 2023 Gallup survey found that 75% of American adults think AI will reduce the total number of jobs in the next ten years, while only 6% think it will increase the total number of jobs and 19% believe it will have no impact. However, a different Gallup survey conducted the same year found that only 22% were worried their job would become obsolete because of technology.

Major Themes

Through our interviews and convening with experts and stakeholders, we identified several opportunities and concerns related to AI’s potential impacts on the workforce. The following section summarizes these key issues by theme.

Managing Job Disruption

Researchers have attempted to quantify the impacts that AI may have on various occupations and tasks, but there remains a high level of uncertainty in these projections, which have already shifted as new information emerges and technologies evolve. Virtually all participants agreed that AI is likely to lead to at least some job disruption, but the certainty and predicted extent of this disruption was heavily debated. Some pointed out that white-collar occupations are more exposed to AI than other kinds of jobs, arguing the forthcoming wave of workforce disruption may be different than past ones.

Additionally, more experts expressed concerns about short-term job disruption due to AI automation compared to potential long-term impacts on employment. Many mentioned how historically, new jobs have been created as new technologies emerged to make up for job losses, while other jobs were modified to enable workers to remain in the same job but perform new tasks in the place of those that became automated. The group expressed a general consensus around the need to train workers for the jobs of the future as they emerge, but some experts raised the need to proactively guide policy and technology to create new tasks that could form the basis for entirely new jobs. Additionally, widespread job market disruption raised discussion of larger impacts on the nature of work, including work hours, contract work, and the quality of jobs.

Productivity and Innovation

Across industries, there is excitement around AI’s potential to significantly boost productivity and innovation; however, there is also a considerable amount of debate and disagreement on the topic. Many experts see AI as a tool that could help reverse the decades-long productivity slump in the U.S., likely increasing wages and standards of living as a result. In contrast, some skeptics argue that projected productivity gains from AI are overstated, and absent sufficient policy measures, these gains will not necessarily benefit workers in the form of higher wages, better benefits, or more leisure time.

Preliminary research has found some evidence of productivity gains through AI. In several studies, such as ones looking at AI impacts on taxi driving, call centers, and writing, productivity gains tend to be concentrated among less productive workers as AI tools enabled them to catch up to more productive colleagues completing the same task. But evidence also suggests that AI does not have uniformly positive impacts for productivity, as demonstrated by a study of consultants which found that AI decreased worker productivity on certain tasks (despite improving it in others). Skeptics of AI’s productivity-boosting potential argue that these studies only find productivity gains in a narrow set of tasks within a short-term period. In contrast, optimists argue that since these studies focus on prevailing tasks and occupations, the productivity impacts of AI may be even higher once new business models emerge and AI is more widely integrated across industries and occupations.

Inclusive Growth and Opportunity

Many experts raised the importance of ensuring that advancements in AI provide opportunities for all Americans. Some expressed concerns that entry-level, white-collar jobs are more vulnerable to AI automation, potentially curbing social mobility for certain workers. Additionally, some were concerned that potential productivity gains from AI adoption may not yield benefits for the average worker, and that worker bargaining power could be weakened. However, others posited that AI may reduce inequality by automating much of the work done by highly paid workers or fill areas where there is a labor shortage, resulting in more savings and opportunities for low and middle-income workers. For example, an economist mentioned how the automation of some tasks performed by a radiologist could result in lower healthcare costs for the average worker and allow nurse practitioners, with the assistance of AI tools, to take over some of the work radiologists would have done otherwise. A labor economist even argued that there is a possibility that AI could help rebuild the middle class if the right policy measures are put in place.

Workplace Governance

The use of AI in the workplace raises important questions about workplace governance, including issues of bias, surveillance, and safety. For example, experts raised concerns about employees’ data rights, pointing to claims that employers are using workers’ data to train AI models without adequately compensating those workers. Others pointed to regulatory uncertainty and a lack of clear and consistent rules as a challenge to AI governance. More optimistically, an expert highlighted how AI could help alleviate certain workplace governance issues. They argued that while AI systems can exhibit bias, AI bias is easier to identify and fix than human bias, which creates opportunities to integrate AI-powered systems and programs in the workplace that are less biased than their human counterparts.

Areas for Further Exploration

Our discussions with experts and stakeholders yielded several areas ripe for further exploration.

Workforce Training

AI is expected to disrupt, create, and modify jobs across industries, impacting demand for various skills. Stakeholders and experts agreed that more needs to be done to improve workforce training and help workers thrive in a rapidly evolving economy, with many emphasizing the importance of lifelong learning, innovative credentialing, and AI literacy. Some further pointed to corporate and civil society initiatives, while others mentioned government initiatives as important parts of the solution.

Social Safety Net Programs

Social safety net programs are often designed to help workers experiencing job loss or shocks to their income. Many experts argued that social safety net programs should be reassessed to evaluate whether they are making the best use of their resources and adequately designed to accommodate potential job disruption from AI. Such a review could look at the efficacy of current spending and assess whether widespread technological disruption may necessitate changes to these programs.

Workplace Governance Mechanisms

As discussed, there is a need to address governance challenges likely to arise from AI adoption in the workplace. Potential mechanisms range from voluntary standards and best practices to new laws and strict regulations. The National Institute of Standards and Technology’s (NIST) Risk Management Framework is an example that provides guidance for employers to build upon as AI applications are integrated into the workplace. Some experts suggest leveraging existing laws to address workplace governance, such as revisiting current labor laws and regulations to identify areas in need of modernization. Some emphasized the importance of collective bargaining and unions for driving AI governance standards, while others emphasized how setting clear rules can facilitate responsible AI adoption.

Research and Human-Complementary AI

Experts and stakeholders expressed the need for more research to inform and address the above challenges to AI adoption in the workforce. This includes research on AI’s potential impacts for jobs and worker productivity as well as to explore privacy-preserving and de-biasing techniques. A few experts mentioned the role of further research in developing AI systems that complement humans’ skills, or “human-complementary AI,” to drive AI technologies toward job creation and increase demand for workers. Others have expressed skepticism about the effectiveness of such R&D policies.


Establishing appropriate infrastructure around AI—including computing resources, broadband, and high-quality data—and its adoption in the workplace could help boost productivity and innovation. Experts flagged the National Artificial Intelligence Research Resource (NAIRR) as an important measure that should be nurtured in the burgeoning AI era.

Other Approaches

Additional areas stakeholders and experts identified as ripe for further exploration include:

  • Finding pathways to help the federal workforce adopt AI and build expertise
  • Identifying feasible AI applications for education and workforce training
  • Examining provisions in the tax code to encourage more worker augmentation


AI is already impacting the U.S. workforce, but the full extent of AI’s influence will continue to unfold as more businesses experiment with and adopt AI technologies. The rapid proliferation of AI raises important policy issues that policymakers and other stakeholders can—and should—proactively address, including managing job disruption, boosting productivity and innovation, promoting inclusive growth and opportunity, and modernizing workplace governance. While discussions around how to best maximize the potential of AI for the workforce while minimizing harmful effects are still early, policies should be informed by empirical evidence and pragmatic engagement by relevant policymakers and stakeholders.

[1] To encourage candor, BPC’s conversations followed the Chatham House Rule, so we avoid attributing what people said to specific individuals throughout this document. However, we sometimes cite publicly available work that was referenced in our conversations (and not necessarily cited by the author of the work referenced).

Read Next

Support Research Like This

With your support, BPC can continue to fund important research like this by combining the best ideas from both parties to promote health, security, and opportunity for all Americans.

Give Now