How exposed is North Carolina's workforce to artificial intelligence (AI), and where in the state is that exposure concentrated? In LEAD's first statewide look at these questions, we pair one of the most widely cited exposure measures (Eloundou et al., 2024) with North Carolina employment data1 to show which occupations are most exposed and how much of the workforce they represent. Exposure is not a forecast of job losses, but knowing where the potential for AI use is greatest can help policymakers and workforce leaders target training and other support as adoption grows.
People are also reading…
Key takeaways
- About a quarter (26%) of North Carolina's workforce is in highly exposed occupations, slightly above the nation’s share (24%).
- Office and administrative support occupations make up the largest share of the highly exposed workforce (34%), followed by business and financial operations (22%) and computer and mathematical (16%) occupations.
- Highly exposed occupations are most concentrated in the Triangle and Charlotte but represent at least a sixth of the workforce in every region.
- High exposure signals potential for AI use, not job losses. AI's impact on the labor market appears small so far, and nearly all North Carolina businesses using AI report no change in employment.
What the AI exposure scores measure
Occupational tasks are rated on whether a large language model such as ChatGPT, on its own or with additional software, could cut the time needed by at least half without lowering quality (Eloundou et al., 2024). An occupation's exposure score combines these ratings across all of its tasks, so a higher score means more of the occupation’s work could be done faster with AI.2 For example, software developers have an exposure score of 87%. Nearly all of their tasks are rated exposed, from developing and modifying software to testing systems. One exception is training users on new or modified equipment.
Two points are important when interpreting exposure scores. First, they measure the potential for AI use, not current adoption. So far, observed AI use lags well behind what is technically possible in every occupation group (Massenkoff & McCrory, 2026; OpenAI, 2026).
Second, exposure does not indicate whether jobs will be lost. To date, AI's impact on the labor market appears small. A recent Stanford review (Mahoney et al., 2026) reports little evidence that AI is reducing employment in AI-exposed occupations, apart from early signs that it may be weakening demand for young white-collar workers. In North Carolina, about 97% of businesses using AI reported no change in employment in the Census Bureau's Business Trends and Outlook Survey (see our previous analysis), and 95% reported no change in the past year in LEAD's 2026 Employer Needs Survey.
About a quarter of North Carolina's workforce is in highly exposed occupations
To measure how much of the workforce is in highly exposed occupations, we ranked U.S. occupations by exposure score and divided them into quintiles, five groups with about the same number of occupations each. Because occupations differ in size, a quintile's share of the workforce can be more or less than 20%. Occupations in the most-exposed quintile have exposure scores of 51.2% or higher, meaning AI has the potential to speed up more than half of their work tasks. We refer to these occupations as highly exposed.
In North Carolina, 26% of the workforce, or about 1.25 million jobs, is in the most-exposed quintile, slightly above the nation’s share of 24%. Across all five quintiles, North Carolina's distribution is broadly similar to the nation's.
Office and administrative support occupations account for the largest share of the highly exposed workforce
By occupation group, office and administrative support accounts for about a third (34%) of the highly exposed workforce, followed by business and financial operations (22%), computer and mathematical (16%) and sales and related (14%). The largest individual occupations in this quintile are customer service representatives and office clerks, both in office and administrative support, followed by software developers. In contrast, the least-exposed quintile is made up almost entirely of occupations centered on physical tasks, led by hand laborers and material movers, fast food cooks, and janitors and cleaners.
Average wages are generally higher in more exposed quintiles. However, the most-exposed quintile ($78,500) has a lower average than the fourth ($102,200) because it includes large, lower-paying office and administrative support occupations.
Exposure is highest in the Triangle and Charlotte but reaches every region
Highly exposed occupations make up between 17% and 35% of the workforce in North Carolina's regions. Durham-Chapel Hill (35%), Raleigh (30%) and Charlotte (27%) are the only regions above the statewide share of 26%, while the three nonmetro regions and the Goldsboro and Hickory metro areas have the lowest shares. Highly exposed office and administrative support occupations make up a similar share everywhere, 8% to 11% of employment in every region. Most of the difference between regions comes from computer and mathematical and business and financial operations occupations. For example, highly exposed computer and mathematical occupations range from less than 1% of employment in Goldsboro to 8% in Durham-Chapel Hill.
Conclusion
AI exposure in North Carolina is not limited to computer and math jobs. It extends to office work of all kinds, from customer service to accounting, in every region of the state. How AI will affect this work remains uncertain. For example, it could alter how some tasks are done, speed up others or change demand for some roles. Knowing where exposure is highest can help policymakers and workforce leaders prepare for these possibilities. That may mean training workers to use new tools, updating education programs or helping workers move into different roles.
As adoption grows, LEAD will continue to track the latest data for early signs of AI’s impact on North Carolina's workforce. Our next post will examine how well workers in highly exposed occupations are positioned to adapt to potential AI-related changes.
- 1
Employment estimates are from Occupational Employment and Wage Statistics (OEWS), May 2025, which cover wage and salary workers in nonfarm establishments and exclude the self-employed.
- 2
Exposure scores are Eloundou et al.'s beta (β) measure, based on GPT-4 ratings. A task counts fully if a large language model alone could speed it up, and half if additional software is needed. Core tasks are weighted twice as much as supplemental tasks.
