What was studied
OpenAI Economic Research published the first Work at the Frontier report, analyzing more than 800,000 work-related messages from U.S. ChatGPT users. The researchers separated generic activities such as writing, summarizing, and scheduling, then classified whether the remaining work belonged inside or outside the user's occupation.
Key findings
- 16.8% of all work-related messages were associated with another occupation.
- After generic work was removed, 43.5% of occupation-specific messages fell outside the user's occupation.
- Among average users, the outside-occupation share was 18.9% in workspaces with two to five seats and 16.3% in workspaces with more than 100 seats.
- Financial calculation and technology troubleshooting repeatedly appeared as crossover tasks across occupations.
Why it matters
AI may change work without eliminating an entire role. A marketer can troubleshoot a website, or a small-business owner can perform basic financial analysis that once required a handoff. General-purpose AI may therefore matter most where specialist resources are scarce.
Limitations
This is an observational study of ChatGPT usage. Message classifications do not directly prove that work was completed or productivity improved, and the sample does not represent workers who do not use ChatGPT. The workspace-size pattern is also not evidence that AI caused the organizational difference.