OpenAI reveals how AI expands jobs

An OpenAI study shows that workers are not only using AI to perform their usual tasks better. They are also turning to it for activities traditionally associated with other professions. Some of those tasks stop being one-off experiments and start becoming part of their regular work.

The analysis reviewed more than 1.5 million work-related messages in ChatGPT between April and July 2026. The research is part of the Work at the Frontier report, which examines how AI is changing tasks before job titles change.

From trying a task to making it part of the job

OpenAI calls this phenomenon task crossover: for example, someone in a sales role using AI to draft advertising materials, or a professional adding analysis from another field to their routine.

Among approximately 6,200 continuously observed workers, tasks outside their occupation that they had already used rose from 13.1% of their occupation-specific AI activity in April to 25.9% in July. The figure does not show that all these workers officially changed jobs, but it does suggest that some new activities are repeated and integrated into their workflows.

The likelihood of returning to a task was also higher when the person had used it the previous month. In comparable observations, workers repeated a task from another occupation 23.6% of the time, compared with 8.4% among those who had not used it the previous month.

Repetition varies widely by activity type:

• Talking with customers about goods or services: 54% returned to the task the following month.
• Writing advertising or promotional copy: 44%.
• Creating marketing materials: 37%.
• Explaining financial information: 15%.

The average return rate for cross-occupation tasks was 18.5%. The differences may be related to how easily each task fits into a routine, company policies, or the risk of making a mistake.

How workers ask AI for help

The study also identifies a difference in how people write their instructions. When a task falls outside their usual occupation, workers tend to use shorter prompts and ask for fewer explanations, tutorials, specific formats, or recommendations.

Instead, they are more likely to provide examples, documents, or context, and to ask AI to review or verify something. One possible explanation is that they are not trying to learn an entire profession from scratch. They bring a real problem and some of the necessary information, then use AI to apply knowledge from another field.

For example, someone might provide a company document and ask for help turning it into a marketing asset, checking a financial interpretation, or preparing for a conversation with a customer. In this context, AI provides access to knowledge that was previously more separated between departments.

What this changes for you and for companies

The finding suggests that jobs can expand before their titles change. A worker starts by trying a new task, discovers that AI makes it possible to handle, and eventually adds it to their routine. Over time, the division of labor within a company can change without a new role appearing on the organizational chart.

For companies, this means that buying access to an AI tool is not enough. It also matters to redesign processes, clarify who can take on new tasks, and establish controls for reviewing results, especially in areas where a mistake has serious consequences.

For you, the change may appear in the mix of activities that make up your job: fewer boundaries between functions and more hybrid work. The question is not only what tasks AI can perform, but which ones you start doing because they are now more accessible.

The study offers an early snapshot, not a definitive prediction about employment. What remains to be seen is whether these cross-occupation tasks become recognized responsibilities, whether they raise expectations for existing roles, and how companies respond through training, oversight, and new ways of dividing work.