ChatGPT at work is no longer just a faster way to draft an email. After analyzing more than 800,000 professional messages, OpenAI finds that employees increasingly use it to handle tasks that used to belong to another job entirely, to the point where formal role descriptions are starting to blur.
Key Takeaways
- 16.8% of work-related messages deal with a task tied to a different occupation than the user’s own.
- On strictly occupation-specific messages, that crossover rate climbs to 43.5%, peaking in customer support, design and HR.
- OpenAI frames it as job boundaries becoming more flexible, a shift that may show up before official labor statistics catch it.
Have an AI Sum Up This Article
ChatGPTNearly Half of Job-Specific Prompts Reach Into Someone Else’s Role
The finding comes from an analysis of more than 800,000 messages sent by US ChatGPT users, a study laid out in OpenAI’s own report on how AI is expanding what people do at work. The headline figure is the crossover rate: the share of prompts that deal with a task belonging to an occupation other than the user’s.
Across all work-related messages, that rate sits at 16.8%. But when the sample is narrowed to strictly occupation-specific prompts, the ones tied to a defined skill, the crossover jumps to 43.5%. Put plainly, nearly one professional request in two has ChatGPT at work on ground that is not the user’s job description.
The examples are concrete: contract review, data analysis, website troubleshooting. Tasks long reserved for a lawyer, an analyst or a developer, now attempted directly by people who hold neither the title nor the training. OpenAI calls the pattern task crossover.
The shift lands in a moment where the chat assistant has become an office reflex. Mass adoption is real, with Gemini now closing in on a billion monthly users, and that normalization feeds the crossover OpenAI measures.
Who Wins and Who Loses When Versatility Becomes the Default
For users, the consequence is immediate. With ChatGPT at work, an employee can now produce a first-draft contract, a numbers-driven analysis or a technical fix without routing it through the in-house specialist. Versatility is no longer a vague line on a resume, it becomes a daily practice, along with the quality and control risks that come with it.
That move is amplified by tools that no longer just answer. When ChatGPT can now drive an entire desktop by voice, the line between asking a question and delegating a full task fades further. Crossover of tasks becomes crossover of execution.
On the competitive and organizational side, the pressure changes shape. The highest crossover rates hit customer support, at 77%, design, at 75%, and human resources, at 69%, ahead of legal at 56% and marketing at 53%. The most exposed roles are the ones whose work leans on repeatable tasks a model absorbs with ease. None of these teams asked for ChatGPT at work to redraw their remit, yet the figures show it happening from the bottom up, one prompt at a time, long before any manager signs off on a reorganization.
For leadership, that reframes hiring. If one profile already covers part of another role’s work through AI, the watertight job description falls apart. OpenAI’s chief economist puts it bluntly: the boundaries between jobs are likely already becoming more flexible because of AI. The signal runs ahead of official employment data, which will take months to register it. For rival labs, the takeaway is just as sharp: the assistant that best handles a lawyer’s, a designer’s and an analyst’s work in one interface wins the office, and that is now the battleground.
More articles on Horizon
- Claude shared chats got indexed by Google
- Apple delays its AI glasses over privacy concerns
- Anthropic’s Open-Weights Position, Explained
What the End of Watertight Roles Means for Your Job
This broad crossover is not a statistical curiosity, it is the same underlying move as the white-collar threat from AI agents the UN has already flagged. When part of your value came from being the only one who could review a contract or clean a dataset, that edge erodes the moment a colleague does it in three prompts.
The real risk is not being replaced by a machine overnight. It is subtler: watching your role dissolve into an assisted versatility where everyone nibbles at everyone else’s turf, until an organization decides it needs fewer specialists. Mid-level roles, the ones built mostly on repeatable tasks, are first in line. If you want to understand exactly where your own position sits in that shift, this free report maps how exposed your role is and what to build next.
The counter-move is a shift in where your value sits. What stays scarce is final judgment, ownership of the decision, the ability to frame a problem and validate what the tool produces. Driving AI into someone else’s field is within everyone’s reach; guaranteeing the quality of what comes out is far less so. That is the layer to build while the boundaries are still moving.
Follow the story on Horizon.
