Job titles might stay static, but the actual work inside them is mutating fast. New data from OpenAI’s Economic Research team reveals that professionals aren’t just dabbling in AI; they are systematically adopting tasks historically reserved for other departments. Analyzing 1.5 million ChatGPT messages between April and July 2026, researchers found that cross-occupation AI use isn’t a fleeting experiment. It is sticking. For the 6,200 workers tracked consistently, these boundary-crossing tasks grew from 13.1% of their AI activity in April to 25.9% by July. This isn’t random exploration. It is the early architecture of role expansion.
The way people prompt reveals their confidence levels. When workers stick to their core expertise, they ask for explanations and specific formats. But when they step outside their lane, the behavior shifts. They write shorter prompts. They stop asking “how” and start asking “check.” Instead of seeking a tutorial, they provide context—documents, examples, background info—and ask the model to apply external knowledge to their specific problem. They are using AI to borrow expertise, not to learn a new trade from scratch. This distinction matters because it suggests AI is lowering the friction of collaboration between silos, allowing a finance analyst to draft marketing copy without needing to become a marketer first.
Recurrence rates highlight which tasks are actually viable for this kind of crossover. Workers returned to customer discussion tasks 54% of the time and promotional writing 44% of the time. In contrast, explaining financial information saw only a 15% return rate. The average next-month return rate for cross-occupation tasks sat at 18.5%, significantly higher than the 8.4% baseline for workers who hadn’t tried the task previously. This gap proves that once workers find utility in stepping out of their defined roles, they tend to keep doing it. Marketing and customer-facing tasks appear to have lower perceived risks or higher immediate rewards, making them prime candidates for AI-assisted role blending.
This trend signals a quiet transformation in organizational design. Companies often focus on buying AI licenses, but the real challenge is redesigning work to accommodate these new, hybrid responsibilities. If an engineer starts regularly handling customer support queries via AI, their job description hasn’t changed, but their output has. Leaders need to recognize that work design is now as critical as tool access. Ignoring this shift means missing how labor is actually being divided in the age of intelligent assistants.
