Is your upcoming colleague going to be artificial intelligence? According to recent findings, for approximately one in five workers, the answer may very well be affirmative. Researchers from Harvard Business School conducted an extensive analysis of over 19,000 job listings spanning more than 900 different occupations, assessing the potential for AI to take over these roles using ChatGPT. They found that around 18% of the jobs studied could potentially be automated, or at least significantly augmented by technology.
The team behind our Get the Facts initiative has developed interactive tools that allow users to explore how various jobs rank in terms of automation potential or search specific roles of interest. The study highlighted customer service, interpretation or translation, and court clerk positions as the most vulnerable to automation, while sectors such as construction and transportation, which involve hands-on tasks, appeared least susceptible to AI takeover. There are also roles in finance, science, and medicine that fall somewhere in between, where AI could assist or enhance current responsibilities.
It’s important to note that this data was derived from job postings as opposed to real-world employment scenarios. Researchers observed a noticeable decline in listings for jobs leaning heavily on repetitive tasks, countered by a 20% increase in demand for positions requiring analytical, technical, and creative expertise.
Curious about how your job measures up in this analysis? Head over to our Get the Facts Data Team's interactive database for a closer look.
With the rise of generative artificial intelligence, such as ChatGPT and Claude, professionals across various fields have expressed concerns about potential job security. The Harvard study aimed to clarify this issue, revealing a 13% decrease in job postings related to highly structured roles that AI could easily perform since 2022. Meanwhile, sectors requiring more complex skills have seen a boom.
The research team leveraged two databases: the Occupational Information Network and U.S. job postings from LightCast, to forecast job market trends from 2019 through March 2025. The Occupational Information Network, which is a U.S. Department of Labor resource, catalogs over 900 occupations and more than 19,000 distinct tasks.
Using ChatGPT, researchers assessed the level of automation possible across these roles, assigning scores based on the likelihood of AI performing specific tasks. A job could also receive a separate score based on the extent to which AI might enhance efficiency within that role. Of the 911 jobs analyzed, roughly 164 received a score of 0.5 or higher for automation potential. Positions such as correspondence clerks topped the list for automation feasibility, followed closely by interpreters, translators, and clerical roles. On the contrary, jobs requiring manual, technical skills—mostly found in production, construction, and transportation—scored zero, indicating that AI is unlikely to replace these roles.
The Get the Facts Data Team has created an interactive tool below, allowing users to gauge the automation feasibility of various jobs. Categories are available for selection, helping you discover the automation scores for specific occupations.
Professions like microbiologists, financial analysts, and clinical neuropsychologists emerged in the study as areas where AI could boost efficiency. For example, while investment managers and analysts can harness AI tools to analyze market trends, human judgment remains essential for certain decision-making processes.
The researchers, Suraj Srinivasan, Wilbur Xinyuan Chen, and Saleh Zakerinia, concluded that AI is fundamentally altering the workforce by shifting labor demand and required skill sets. There’s a downward trend in job postings for jobs necessitating simple skills, but a corresponding increase in the need for advanced competencies.
However, the study has some limitations; it primarily examined short-term impacts, failing to consider the long-term ramifications of AI on employment. The researchers caution that the manner in which businesses deploy AI technologies will ultimately dictate whether the job market experiences growth or contraction. They advocate for companies to invest in employee training programs, helping workers adapt to roles that incorporate AI while still emphasizing skills that cannot be automated. Additionally, the reliance on job postings as a measure of the labor market means that distorted hiring trends could be present, especially concerning "ghost job postings" in white-collar sectors.
