Imagine a small-business owner using ChatGPT to prepare a marketing campaign, troubleshoot software and interpret a financial calculation in the same afternoon. New research from OpenAI suggests this kind of boundary crossing is becoming a defining feature of AI at work. Rather than measuring only which jobs could be automated, the study examined how people use ChatGPT to perform tasks traditionally associated with other occupations.
OpenAI analysed more than 800,000 work-related messages from users in the United States. It reported that 16.8 per cent of all work messages crossed occupational boundaries. When common activities such as routine writing and scheduling were removed, 43.5 per cent of occupation-specific requests involved tasks normally linked to another profession. People were using AI to reach into areas such as marketing, software support, financial analysis and regulatory explanation.
The pattern was especially visible in smaller businesses. OpenAI found that nearly 19 per cent of work-related requests at the smallest organisations crossed job boundaries, compared with about 16 per cent at larger firms. That result makes intuitive sense: a lean team cannot employ a specialist for every problem. An accessible AI assistant can help one person explore unfamiliar work, create a first draft or understand enough to ask an expert a better question.
However, task crossover is not proof that jobs are disappearing or that users are suddenly qualified professionals. The research cannot determine whether AI created new responsibilities or merely helped people perform duties they already had. A generated financial model may still require an accountant, regulatory guidance may require a lawyer and technical changes may need an experienced engineer. AI can lower the barrier to beginning a task without removing the need for accountability.
For employers, the immediate change may be broader roles rather than smaller headcounts. Training should focus on verification, judgement and knowing when to escalate work to a specialist. Job descriptions may increasingly value people who can combine domain expertise with AI-assisted skills from neighbouring fields. The most productive organisations will not ask only which tasks can be automated. They will ask which employees can now attempt more valuable work, while keeping clear boundaries around decisions that still require qualified human expertise.