ChatGPT task crossover rises in OpenAI analysis of work use

OpenAI says ChatGPT task crossover is becoming a measurable pattern at work: in an analysis of more than 800,000 work-related ChatGPT messages, 43.5% of job-specific prompts were tied to a profession other than the user’s own. The company said the pattern shows workers using AI to take on tasks that would previously have been more likely to sit with specialists.

OpenAI calls the behavior “task crossover.” The analysis looked at work-related ChatGPT messages and categorized occupational tasks using O*NET, the U.S. occupational database that maps work activities to standardized job profiles. OpenAI said it left out broad productivity uses such as writing, summarizing and scheduling, which would otherwise blur the distinction between general office work and occupation-specific work.

What is ChatGPT task crossover?

ChatGPT task crossover is OpenAI’s term for cases where a worker uses the chatbot for a task associated with a different occupation. In practice, that can mean a non-lawyer asking for help reviewing a contract, a non-analyst using ChatGPT for data analysis, or a non-engineer trying to address website problems.

The company said marketing and engineering were the areas where crossover appeared most often. Those categories are broad, but they are also common pressure points inside companies where demand can exceed available staff, especially outside large enterprises with dedicated teams.

Where is OpenAI seeing the strongest effect?

OpenAI said the pattern is more pronounced at smaller companies. Its explanation is straightforward: smaller employers often lack specialist departments, so workers are more likely to stretch across functions. In that setting, ChatGPT becomes a tool for filling gaps in marketing, technical support, analysis or legal-adjacent work, though OpenAI’s summary did not claim those users are replacing trained professionals outright.

For operators, the finding is a useful signal because it focuses less on AI as a standalone product category and more on how employees are already changing workflows. A founder may not hire a full marketing function early, but employees can still use AI systems to draft campaign material, troubleshoot web issues or prepare analytical work that might previously have required outside help.

What does this say about jobs?

OpenAI described the usage data as an early indication that job roles may be changing before formal job titles and descriptions reflect that change. That is a narrower claim than saying AI has already redrawn org charts. The evidence cited is based on ChatGPT usage patterns, not on hiring data, compensation changes or revised job classifications inside companies.

The distinction matters for employers selling or buying workplace AI. Usage logs can show which tasks workers try to push into AI systems, but they do not by themselves show whether the output was accurate, approved, compliant or useful in production. Contract review, data analysis and website troubleshooting all carry different levels of risk depending on the company and the person using the tool.

Still, the 43.5% figure gives a concrete measure to a trend many software vendors have been pitching in vague terms: AI tools are not only speeding up existing work, according to OpenAI’s analysis, they are also letting workers reach into adjacent functions. The next question for companies is whether that activity becomes formal process, shadow work, or a reason to redefine what a given role is expected to cover.

This story draws on original reporting from The Decoder.