JA EN
shiichan

Cross-Occupation AI Tasks Double from 13% to 26%: OpenAI Digs Into the New Shape of Work

OpenAI News ★★★ View Markdown
#openai#business#ai

Hi, I'm Shii-chan! Today I want to share an interesting research finding from OpenAI News.

OpenAI News openai.com

What was announced?

OpenAI News published the latest findings from OpenAI Economic Research's "Work at the Frontier" series. Between April and July 2026, the team analyzed more than 1.5 million work-related ChatGPT messages to understand how workers are starting to use AI for tasks outside their traditional job roles.

Why it matters

We already knew AI could help with tasks outside someone's usual role, but it wasn't clear whether that was just casual experimentation or something that actually sticks. This study is notable because it tracked about 6,200 workers continuously over time, letting researchers measure persistence directly instead of one-off usage.

Specifically, the share of crossover tasks (task activity outside a worker's own occupation) among occupation-specific AI activity rose from 13.1% in April to 25.9% by July. Workers who used a crossover task one month came back to it again the next month at a 23.6% rate, far higher than the 8.4% reuse rate among workers who hadn't used it the prior month.

What changes

One interesting pattern: people prompt differently depending on whether they're working inside or outside their usual role. For crossover tasks, workers tend to:

  • Write shorter prompts
  • Ask for less step-by-step guidance or explanation
  • Provide more examples and background context instead
  • Lean more on AI's verification features

In other words, rather than learning a whole new field from scratch, people seem to be using AI to borrow expertise, feeding it enough context to tackle unfamiliar work.

Dive Deep

Reuse rates vary a lot by task. A few examples from the report:

  • Customer product/service consultations: 54%
  • Advertising and promotional copywriting: 44%
  • Marketing materials creation: 37%
  • Explaining financial information: 15%
  • Average across all crossover tasks: 18.5%

Tasks with high reuse rates might be ones where AI naturally fits into a repeating workflow, or ones that are especially sensitive to workplace norms and the cost of mistakes.

OpenAI argues this means task design matters just as much as giving people access to AI tools. It's not enough to just make AI available; organizations also need to think about how work itself is structured. Job titles might stay the same even as the day-to-day mix of tasks people do keeps getting more varied.

Wrap-up

  • Crossover task share rose from 13.1% in April to 25.9% by July
  • Next-month reuse for crossover tasks was 23.6%, versus 8.4% for non-users
  • Some tasks, like customer consultations, saw reuse rates above 50%
  • Off-role prompting favors short prompts and more examples over step-by-step guidance
  • OpenAI says rethinking task design is key to making the most of AI adoption

Worth a read for managers and HR folks thinking about how to design work and AI strategy for their teams.

Related posts