OpenAI said workers in jobs more vulnerable to automation are already using ChatGPT far more than everyone else.
In a new study, the company said occupations it classifies as more susceptible to automation use ChatGPT about three times more than jobs facing less immediate exposure. Roles such as data-entry keyers, bookkeepers, and similar administrative work ranked among the most exposed categories, according to the report.
The report introduced what OpenAI described as an “AI jobs transition framework,” a model that groups more than 900 occupations based on how much of the work AI can realistically handle today, how essential humans remain in those roles, how demand shifts when costs fall, and how AI is actually being used on platforms like ChatGPT.
According to the findings, 18% of workers are in roles facing the highest near-term automation risk. These include jobs such as data entry, bookkeeping, and customer service.
Another 24% of roles could see employment shrink even though the work itself is still expected to remain human-led. OpenAI pointed to occupations such as HR specialists as examples where demand could fall without full automation.
The report also said 12% of jobs could see employment expand because of AI adoption. Software developers were listed as one example, reflecting how AI can increase output and create new demand rather than only replace labor.
Meanwhile, 46% of occupations face the least immediate threat of disruption, according to the study. Roles such as teachers and home-health aides were cited as work that still depends heavily on human judgment, care, and in-person interaction.
OpenAI Chief Economist Ronnie Chatterji said AI capabilities have increased at “unprecedented speed,” adding that early adopters are already seeing meaningful productivity gains. He noted that while the labor market impact is still uneven and sometimes contradictory, the pace of progress means broader disruption remains possible.
He also said the report introduces a framework designed to better measure how AI affects jobs by combining capability, human necessity, cost-driven demand shifts, and real-world ChatGPT usage data. He added that the goal is to help policymakers and businesses understand which jobs face pressure first, which are likely to be redesigned, and which could grow over time.
The report said clear signs of mass disruption have not yet appeared in unemployment data, suggesting the transition may be slower and more uneven than simple replacement narratives suggest.
The bigger question, according to the report, is what happens when AI makes tasks cheaper and faster. Companies could reduce headcount, but they could also expand output, launch new services, or shift workers into higher-value tasks. This tension between automation and growth may decide whether AI cuts jobs or simply reshapes them.
David Adubiina
