New research from OpenAI Economic Research examines how workers use AI (specifically ChatGPT) to perform tasks outside their formal occupations and whether those activities become recurring parts of their workflows. The study analyzed over 1.5 million work-related ChatGPT messages from April through July 2026 and finds evidence that some cross-occupation AI use is repeated and increases over time.
Data and methods
- The analysis is based on more than 1.5 million work-related ChatGPT messages collected between April and July 2026.
- Occupations were identified using role or department information provided during ChatGPT Business onboarding.
- The sample excludes training-disabled data and messages that could not be classified.
- Messages were aggregated and anonymized for analysis; researchers did not read individual user messages.
Key findings
- Building on an earlier Work at the Frontier report that documented "task crossover"—workers using AI for tasks typically associated with other occupations—this study asks whether those cross-occupation activities persist.
- Among roughly 6,200 workers who were observed consistently from April through July, the share of cross-occupation tasks within occupation-specific AI activity rose from 13.1% in April to 25.9% in July. This pattern is consistent with some cross-occupation assistance becoming integrated into ongoing workflows rather than remaining a one-off experiment.
- In a separate matched one-month follow-up analysis, workers returned to a cross-occupation task used in the previous month 23.6% of the time, compared with 8.4% recurrence for comparable workers who had no observed use of that task in the prior month. Similar gaps were present for within-occupation and general tasks.
- Recurrence varies substantially by task type. The report notes that some cross-occupation tasks showed relatively high month-to-month return rates, while others were lower. For example, returning to the task of explaining financial information occurred about 15% of the time. The average next-month return rate across cross-occupation tasks was 18.5%.
How AI is used differently for inside-versus-outside tasks
- Workers prompt AI differently depending on whether a task fits within their usual role. On average, prompts are shorter when the task lies outside their occupation.
- For outside-occupation tasks, workers are less likely to request explanations, step-by-step guidance, a specific output format, or explicit advice.
- At the same time, they are more likely to provide examples or background information and more likely to ask the AI to check or verify something.
- These patterns suggest workers are often using AI to "borrow" expertise: rather than asking the AI to teach an entirely new field, they bring a problem and relevant context (a document, example, or colleague’s input) and ask the AI to apply knowledge from another domain.
Why this matters
The findings sketch a pathway by which AI could reshape job content before job titles change. A worker experiments with a task traditionally associated with another occupation, finds AI useful for that task, and begins to return to it as part of regular work. If such activities become routine responsibilities, the mix of tasks within a job could broaden even while the job title remains the same.
OpenAI emphasizes that access to AI tools alone is not the only determinant of how work changes: work design and the arrangement of tasks also matter for how organizations adopt AI and reduce friction between identifying problems and progressing work with AI assistance.
Methodological and privacy notes
Occupations were inferred from role or department information supplied during ChatGPT Business onboarding. The messages analyzed come from a sample of users’ conversations; training-disabled content and unclassifiable messages were excluded. All messages were aggregated and anonymized for analysis, and researchers did not read individual user messages.
Next steps
OpenAI plans to continue studying these dynamics to clarify how AI is altering the division of labor and what that implies for workers, firms, and the broader economy. Important open questions include which specific tasks are most likely to become persistent parts of a job and what training, organizational, or regulatory responses may be appropriate.
