OpenAI Economic Research’s new Work at the Frontier series examines how AI is enabling workers to perform tasks beyond traditional occupational boundaries. The analysis is based on more than 800,000 ChatGPT messages from U.S. users.
Key findings
- 16.8% of work-related messages overall involve AI use, and among non-generic, occupation-specific messages, 43.5% concern tasks typically associated with a different occupation.
- The report calls this phenomenon task crossover: tasks historically linked to one occupation appearing in the AI use of people in another occupation.
What the numbers indicate
These usage patterns suggest AI is changing both how work is done and who does it. Examples include a small-business owner drafting marketing copy or performing basic financial analysis independently; a salesperson using AI to explore customer data that might previously have gone to an analyst; or a marketer troubleshooting a website without waiting for a developer.
Method: separating generic and occupation-specific tasks
Researchers first distinguished broadly shared, generic activities—such as writing, summarizing, and scheduling—that are too common across occupations to count as evidence of crossover. For the remaining non-generic messages, they evaluated whether the task fell inside or outside the user’s own occupation. Among those non-generic messages, 43.5% were outside the user’s occupation, offering an early view of how AI may be reshaping task content before formal job descriptions or titles change.
Patterns across occupations
Task crossover is uneven across occupations and specific tasks. Notable points from the analysis include:
- Design: About 35.2% of messages from designers involve work usually associated with other occupations, while design tasks make up only 1.7% of messages from workers in other fields. Designers therefore draw heavily on outside tasks, but design work rarely appears elsewhere.
- Engineering: 18.5% of engineering messages include tasks from other fields, while engineering tasks account for 7.4% of messages among non-engineers. This suggests engineering work is a significant source of tasks that others take on, such as troubleshooting technical systems.
- Marketing: Marketing shows bidirectional movement. Marketing workers devote 24.3% of their messages to tasks associated with other occupations, and marketing tasks account for 8.9% of messages among workers in other fields—the largest outward share in the sample.
Two directions of crossover
The report identifies two broad patterns: some occupations import many tasks from other fields (e.g., design), while others export tasks that appear across many jobs (e.g., engineering). Marketing operates strongly in both directions, combining tasks from multiple domains and spreading marketing work across organizational roles.
Role of firm size
The size and structure of a business shape how AI alters work. In large firms, employees often have access to specialized teams and established internal workflows, while in smaller organizations the worker closest to the problem is likelier to solve it rather than hand it off. Among average users, the share of outside-occupation tasks falls from 18.9% in workspaces with 2–5 seats to 16.3% in workspaces with more than 100 seats. Among the heaviest users, this monotonic pattern is not observed.
Interpretation and implications
AI usage data provide an early indicator of where work is shifting by revealing how workers experiment with new combinations of tasks before firms rewrite job descriptions or create new job titles. The report draws on the AI Jobs Transition Framework to argue many jobs are likely to reorganize as daily task mixes change substantially. Tracking these usage patterns can therefore inform policy and workplace practice by signaling occupational change earlier than conventional labor-market statistics.
