OpenAI research says ChatGPT work use is crossing traditional job boundaries

OpenAI published new economic research on July 27 arguing that workplace use of ChatGPT is not only automating tasks inside existing jobs, but also helping workers take on assignments that have traditionally belonged to other roles.

The company introduced the first report in a new Work at the Frontier series, built from an analysis of more than 800,000 messages from U.S. ChatGPT users. OpenAI says 16.8% of work-related messages and 43.5% of occupation-specific messages involved tasks associated with another occupation, a pattern it calls task crossover.

What OpenAI found

The report separates broadly shared activities, such as writing or summarizing, from tasks more closely tied to specific occupations. Among those non-generic messages, OpenAI says nearly half crossed occupational lines. The pattern was especially visible in customer experience, design, human resources, legal and marketing work, where users frequently asked for help with tasks normally linked to other functions.

OpenAI frames the finding as an early signal that AI could reorganize jobs before those changes show up in job descriptions, hiring data or formal titles. The examples are practical rather than futuristic: a small-business owner drafting marketing copy or reviewing a contract, a salesperson exploring customer data, or a marketer troubleshooting a website instead of waiting for a developer.

Why it matters

The most interesting detail is not that ChatGPT helps with office tasks, but that the mix of tasks is changing. OpenAI says marketing and engineering work travel widely across occupations, while small organizations appear especially likely to use AI as a generalist tool when specialized support is scarce.

For employers, that points to a different kind of AI adoption question. The issue is not simply whether a model can perform a job, but whether teams are already redistributing work around model-assisted employees. That could alter training, internal controls, software permissions and compliance reviews, particularly when workers begin handling legal, finance or technical tasks outside their original lane.

The research is based on OpenAI's own usage data, so it should be read as a company-produced snapshot rather than a neutral labor-market measure. Still, the scale and specificity make it a useful marker for how AI tools are being absorbed into everyday work: less as a standalone replacement for a role, and more as a way for workers to stretch into adjacent responsibilities.

Sources

• openai.com

Cover photo by Kampus Production on Pexels, used under the Pexels License.