OpenAI task crossover study shows AI blurring workplace roles

OpenAI task crossover research found that 43.5% of occupation-specific work requests made to ChatGPT by U.S. users involved duties usually linked to another profession. The finding, based on more than 800,000 work-related messages, points to a shift in how employees use generative AI to handle work that once moved through specialists or separate departments.

OpenAI published the findings on July 27 in the first installment of its “Work at the Frontier” series. The company described the pattern as “task crossover,” a measure of how often workers use AI for tasks associated with occupations other than their own.

What is OpenAI task crossover?

Task crossover refers to a worker using generative AI for an activity normally associated with a different role, such as a customer support employee running a financial calculation or a human resources employee addressing a technical problem. It does not mean the worker has taken over the full responsibilities of another profession, only that some discrete tasks are moving across job boundaries.

The highest rate appeared in customer experience, where 77% of occupation-specific requests crossed into another field, according to OpenAI. Designers followed at 75%, human resources at 69%, legal at 56% and marketing at 53%.

OpenAI’s data also showed that crossover does not move evenly between functions. Designers used ChatGPT for work outside their field in 35.2% of all their messages, including messages that were not occupation-specific. By contrast, only 1.7% of workers in other fields touched on design tasks, the company said.

Engineering showed a different pattern. OpenAI found that 18.5% of engineering messages involved other fields, the lowest share among the roles it measured. Yet engineering-related work, including troubleshooting, appeared in 7.4% of messages from non-engineers, suggesting that some employees are using AI to resolve technical issues before turning to formal support channels.

Marketing was more open in both directions. Marketers directed 24.3% of their messages toward work tied to other occupations, while marketing tasks appeared in 8.9% of messages from other fields, the highest outward share recorded by OpenAI.

How company size and AI experience affect crossover

Company size had some relationship to the pattern. OpenAI found crossover was higher at smaller workspaces, at 18.9%, than at larger ones, at 16.3%. That difference is consistent with the operating reality of smaller companies, where employees often lack access to a dedicated specialist for every function.

The company also found that among the heaviest users of ChatGPT, workplace size became a much weaker predictor of crossover. The data suggest repeated use may have a stronger connection to role-blending than company structure alone, although OpenAI did not present the findings as proof of a single cause.

Separate research from PYMNTS Intelligence adds evidence on the role of experience. Its latest Consumer AI Benchmark Report found that 61% of workplace generative AI users have used the technology for at least one year. That share rises to 75% among the heaviest users and falls to 42% among light users, PYMNTS reported.

PYMNTS also found that personal users with at least one year of experience complete an average of 11 tasks with generative AI, compared with 6.6 tasks among newer users. In the same report, 31% of users with more than one year of experience described generative AI as essential for managing finances and banking, compared with 13% of newcomers.

OpenAI said its findings do not show that AI is eliminating jobs or that departments are disappearing. The research instead indicates that the practical borders between functions are becoming less fixed as employees use AI to complete parts of work that previously required another role.

This story draws on original reporting from PYMNTS.