Designers pull in far more outside work than they supply, while engineering and marketing tasks travel farthest, turning up often in what workers in other fields write.

More than 800,000 messages written by ChatGPT users in the United States sit behind OpenAI's announcement this week that its own reading of them suggests 16.8% of work-related messages, and 43.5% of the messages that map to a particular occupation, concern tasks belonging to a job other than the one the sender holds.

The company calls that pattern "task crossover" — work that has traditionally belonged to one occupation surfacing when people employed in a different one turn to AI. Its own illustrations: a salesperson can explore a customer dataset that would once have gone to an analyst, a marketer can troubleshoot a website without waiting for a developer, a small-business owner can draft copy and review a contract unaided.

Not every request counts toward it. The method first sets aside as generic the tasks many jobs share, among them writing, summarizing and scheduling. Each remaining message is then tested for whether its task belongs to the job the user actually holds or to some other job. The higher of the two shares applies only to that occupation-specific remainder.

Within that remainder, some fields cross over far more than others: tasks belonging to other jobs account for 77% of the occupation-specific messages from customer experience workers and 75% among designers, falling to 53% among marketers.

Supplying tasks is not the same as borrowing them. Roughly 35.2% of what designers write concerns work normally tied to another job, while design tasks make up just 1.7% of the messages sent by workers in other fields. Engineering runs the other way: its work — fixing software problems, handling technical systems — appears in 7.4% of what other workers send, and marketing's 8.9% is the largest outward share recorded for any job studied.

The report reads that spread as a signal: how people use AI could serve as an early indicator of occupational shifts that standard labor-market figures register only afterwards, because AI lets workers try fresh combinations of tasks before employers redraw job descriptions or mint titles for new ones.