Recent advances in artificial intelligence (AI), and generative AI in particular, have led to an explosion of commercial and public interest in these technologies. Tech firms are investing billions of dollarsicon denoting destination link is offsite to expand their AI capabilities, and millions of Americans are now actively or passively using these technologiesicon denoting destination link is offsite on a daily basis. The latest AI tools are both more powerful and can accomplish a wider range of tasks than their predecessors. Importantly, they are also more accessible. The Large Language Models that power tools like ChatGPT make it possible to interact with these technologies using plain English without the need for advanced programming skills. AI optimists believe it will eventually lead to widespread productivity gains and reshape the future of work (although not all experts share this viewicon denoting destination link is offsite).

What do we know about how recent developments in AI are affecting jobs and the demand for AI-related skills in the labor market? Some of the best systematic evidence comes from online job postings data. The reason is that these technologies have to be developed and implemented by workers with specific skillsets, and they have to be used by workers who are familiar with them. Therefore—to the extent that firms are explicit about what the job entails and the skills they are seeking in prospective employees—job postings provide signals about the demand for AI skills. The fact that most vacancies are now posted online (by some estimates icon denoting Adobe PDF file formaticon denoting destination link is offsite around 60 to 70 percent of all vacancies), combined with constant churn in the labor market, ensures that we can get those signals in real time.

This blog post leverages online job postings data from Lightcast (formerly known as Burning Glass) to document how the demand for AI skills has evolved. Lightcast aggregates job ads from thousands of online job boards on an ongoing basis, thereby capturing the near-universe of online job postings. For each job posting, Lightcast records detailed information, including the employer, job title, industry, location, and any experience or education requirements. Crucially, Lightcast also extracts detailed skills that are listed in the job description. We define "AI jobs" as job postings that list at least one "AI skill," which we identify using Lightcast’s "AI and ML" skill category. This category includes 173 unique skills related to AI and machine learning (ML), and spans general terms like "Artificial Intelligence," specific concepts like "Neural Networks," programming packages like "PyTorch," and applications like "ChatGPT."

While online job postings are a selected subset of all vacancies (some jobs are more likely to be posted online than others), the upshot is that these data represent higher-skill jobs that are more likely to demand AI skills very well. The long-running and high-frequency nature of these data allows us to document changes in the demand for AI skills since 2010 up until August 31, 2024, and the large-scale nature of these data allows us to disaggregate the demand for AI skills by detailed occupations and industries (the data contain more than 400 million unique job postings).

We document three main findings in this blog post. First, the demand for AI skills in the US labor market has been steadily rising since 2010, and this trend seems to have accelerated in the last year. Second, in recent years, the nature of the demand for AI skills has shifted away from ML-related skills toward AI-related skills. Lastly, the demand for AI skills is spreading to a broader set of occupations, industries, and local labor markets.

The demand for AI skills has been rising over time, and this trend seems to be accelerating
Figure 1 plots the share of online job postings for AI jobs. In line with past studies using the same data source (Alekseeva et al. 2021icon denoting destination link is offsite, Acemoglu et al. 2022icon denoting destination link is offsite), we find that the demand for AI skills steadily rose between 2010 and 2019, from 0.16 percent to 0.84 percent. However, extending the series to the present reveals that the pace seems to have quickened in the last year. While demand for AI skills dipped slightly in 2023 (mirroring a postpandemic investment slowdown in the tech sectoricon denoting destination link is offsite), since then it has rebounded in a big way: Over the first eight months of 2024, the share of online job postings for AI jobs was 1.62 percent. Whether this trend will persist is uncertain, but the evidence suggests that the demand for AI skills is accelerating.

There has been a recent shift from ML-related skills toward AI-related skills
Examining the composition of the demand for AI skills reveals some interesting patterns. Although AI and ML are hard to separate, in recent years the demand for AI skills has shifted away from skills one would typically associate with ML toward skills more specifically related to AI. Figure 2 depicts this shift, extracting all AI skills across all AI jobs and plotting the distribution across 4 categories: (1) ML-related skills, (2) AI-related skills (except "Generative AI"), (3) "Generative AI," and (4) all other AI skills. AI-related skills are skills that include the term "AI" (for example, "AI Systems"). ML-related skills are defined analogously but also include "traditional" ML concepts like "Random Forest Algorithm" or "Ensemble Methods" (see the note below figure 2 for more details). All other skills were conservatively assigned to the "other/ambiguous skills" category.

Between 2010 and 2016, ML-related skills made up around 40 to 50 percent of all AI skills listed in online job postings, while the share of AI-related skills was only around 10 to 20 percent. Since then, the balance has shifted. Between 2016 and 2024, the share of AI-related skills has risen to 45 percent, while the share of ML-related skills has declined to 30 percent. The recent emergence of generative AI is also evident in figure 2: Since 2022 and the introduction of ChatGPT, the share of the "Generative AI" skill category has shot up to around 5 percent, underscoring how the field of AI is rapidly evolving as well as the types of AI skills that are in demand as a result.


The demand for AI skills is spreading to a broader set of jobs, industries, and local labor markets
Figure 3 shows the share of occupations, industries, and commuting zones (clusters of counties with strong commuting ties) in which the AI online job posting share is at least 1 percent. Up until 2015, demand for AI skills was highly concentrated. Since then, the demand for AI skills has spread to a larger set of occupations such that by 2024, nearly a quarter of all occupations had some minimal demand for AI skills. Since these occupations are naturally distributed across a more diverse set of industries and locations, the share of industries and commuting zones with some minimal demand for AI skills has also been rising in tandem, reaching 29 and 13 percent in 2024, respectively.

In which jobs has the demand for AI skills gone up? It’s useful to distinguish between computer and mathematical jobs, which are largely responsible for the development and implementation of AI technologies, and all remaining jobs. Figure 4 displays the top 15 occupations in terms of the share of online job postings for AI jobs in 2024, separately for computer and mathematical occupations (panel A) and all other occupations (panel B). The corresponding shares for 2010 and 2017 are shown for comparison. Panel A shows that up until 2017, the demand for AI skills was concentrated in a handful of computer and mathematical occupations, notably computer scientists, data scientists, and statisticians. However, the patterns for 2024 suggest that AI skills are becoming increasingly common in most computer-related jobs. One likely reason is that AI-powered coding tools can help enhance and speed up programming tasks, which all of these jobs involve to varying degrees. Another likely reason is that developing and deploying AI solutions require a team of specialized workers (to set up the necessary data storage and processing infrastructure, network capabilities, security protocols, etc.).

The fact that the demand for AI skills has increased in tech jobs is not surprising. What is striking, however, is that the demand for these skills is rapidly increasing in many other jobs as well. This is because the latest AI technologies can perform a wide variety of tasks (and are increasingly getting better at performing those tasks), including text analysis, pattern detection, speech recognition, image processing, and data analysis. As a result, panel B shows that demand for AI skills is spreading to a wide range of occupations in which the demand for such skills was previously low or virtually nonexistent. These include not only jobs in the life sciences (such as biological scientists, bioengineers, and biomedical engineers) or in business and finance (including economists, financial specialists, management analysts, and financial risk specialists), but also jobs that involve writing (for example, proofreaders, technical writers, writers and authors, and editors).

Overall, the patterns in figure 4 highlight the numerous applications of the most recent AI technologies. These findings are consistent with recent evidenceicon denoting destination link is offsite from a 2023–24 Census Bureau survey in which 27 percent of US firms reported using AI to perform tasks previously done by workers, as well as recent evidenceicon denoting destination link is offsite from a nationally representative household survey showing that generative AI usage is widespread (both at home and in the workplace, and across a wide range of tasks and job categories).

Conclusion
Demand for AI skills is rising and spreading to broader segments of the US labor market. It is still too early to tell whether AI technologies will end up revolutionizing the future of work or not, but the evidence suggests that AI is starting to have a meaningful impact on a growing number of jobs.