Executive Summary
Having access to the right talent is critical to maintaining a competitive edge in artificial intelligence. In the United States, policymakers are actively discussing legislative proposals to grow and cultivate a globally competitive domestic AI workforce. However, little data is available on the U.S. AI workforce and associated talent pipelines outside of the PhD segment.
Yet having access to good workforce data is critical to actually “winning” the competition for AI talent. This brief provides two contributions to better understand the U.S. AI workforce: (1) a definition of the AI workforce based on the government occupational classification system, identifying 54 occupations that either participate or could participate in AI product and application development, and (2) a preliminary assessment and characterization of the supply of AI talent, which consisted of 14 million workers in 2018 (about 9% of total U.S. employment).
Our definition of the AI workforce enables supply-side analysis that is more comprehensive than other commonly used sources, because it is linked to the federal occupation classification system. While many supply-side analyses of the AI workforce rely on sources such as LinkedIn, we use data from the U.S. Census Bureau. Our definition also enables greater analytic consistency across federal government and other datasets that link to this classification system, such as Burning Glass.
Key initial findings regarding the supply of U.S. AI workers include:
- The technical component of the AI workforce struggles with diversity, where a majority of workers are male and not representative in terms of race and ethnicity.
- Four-year college is a common pathway for many AI jobs; however, a sizeable share do not have four-year degrees, particularly in non-technical occupations.
- Degrees in engineering and computer science are among the top fields of study for technical AI occupations; however, non-technical degrees such as business are also common across AI occupations.
- While technical occupations garner much attention, the large number of non-technical occupations in the AI workforce suggests an approach to AI workforce policy that includes a range of education and training pathways.
This brief is the first in a three part series. The second paper will discuss U.S. AI labor market dynamics, while the third paper will provide actionable policy recommendations. Additional future research related to this series will explore topics such as the perceived rise of AI-related certifications and broader manpower and personnel policy implications for the DOD and national security community.