OpenAI Finds 44% of Work ChatGPT Requests Cross Professional Lines

On July 27, 2026, OpenAI's economic research team released a study looking at how its US business customers use ChatGPT at work. The team examined more than a million occupation-related messages sent by workers in different roles. They found that around 44% of these messages asked for help with tasks that usually belong to a different profession. The overlap is strongest between jobs that already share some skills, like marketing and sales, or software engineering and data analysis. But it also shows up in surprising places. Accountants draft marketing copy, support agents help with basic legal questions, and designers write short code snippets. Workers are quietly using the chatbot to do work their official job description does not include. The study suggests that job titles are becoming a less reliable way to plan work. If almost half of the requests cross professional lines, managers may need to redesign teams around tasks rather than around formal roles. Training programs, performance reviews, and hiring plans built only around a single occupation may miss the real shape of the work. OpenAI is not the first to notice this. Earlier surveys from Google and other research groups also found that AI use inside companies is broader but shallower than job titles suggest. Together, these studies are starting to underline a clear shift: in the AI era, what workers actually do is moving faster than what their job titles say they do.