Artificial intelligence is changing who does what at work, not just how it’s done. Can you imagine someone who used to ask for help now solving the task themselves with an AI tool? That’s exactly what OpenAI’s new research in its Work at the Frontier series shows.
What the research found
OpenAI analyzed more than 800,000 ChatGPT user messages in the United States and detected a pattern it calls task crossover: tasks historically associated with one occupation frequently appear in AI use by people from another occupation.
- In general, 16.8% of work-related messages deal with tasks belonging to another occupation.
- If we exclude activities that are too generic (like writing or summarizing), 43.5% of those messages do not match the user’s occupation.
The result? AI enables the first person who encounters a problem to try to solve it without routing the issue through the traditional channel of another team.
Concrete examples and why it matters
Small businesses can write copy, review contracts, or do basic financial analysis without hiring a specialist. A salesperson can use AI to explore a customer dataset instead of waiting for an analyst. A marketer can fix simple website issues without depending on a developer.
This changes the division of labor: fewer handoffs, more experimentation by workers expanding their role. It’s not just efficiency; it’s an implicit reshuffling of tasks inside companies.
Which occupations are changing the most
Among occupations where task crossover is most pronounced, and excluding generic tasks, the presence of tasks outside the occupation is:
- 77% for customer experience workers
- 75% for designers
- 69% for human resources
- 56% for legal
- 53% for marketing
Looking at specific tasks, clear patterns emerge: financial calculations and troubleshooting tech problems are common external tasks across groups. Creating marketing materials appears in five other groups and is especially frequent among design users.
Two directions of change: who receives and who provides
The report describes two distinct movements:
-
Some jobs import many tasks from other occupations. Design is an example: 35.2% of designers’ messages deal with tasks from other roles, but only 1.7% of other people’s work includes design tasks.
-
Other jobs export tasks that show up across many occupations. Engineering illustrates this: only 18.5% of engineers’ messages involve outside tasks, but engineering tasks make up 7.4% of messages in other occupations.
Marketing sits at both ends: it takes many tasks from other fields (24.3%) and its tasks appear widely in other areas (8.9%).
Company size also matters
Organizational structure modulates the phenomenon. In small companies, the person closest to the problem tends to solve it; in large ones, there are more specialized teams and more handoffs.
OpenAI found that the share of tasks outside the occupation falls from 18.9% in workplaces of 2–5 seats to 16.3% in spaces with more than 100 seats. Among the most intensive AI users, the same drop isn’t seen: they may have developed AI workflows that work similarly regardless of company size.
The key signal: AI use reveals changes in tasks before they appear in job descriptions or traditional labor statistics.
What this implies for workers, companies, and policy?
-
For workers: learning to use AI lets you expand what you can do; it’s not about instant replacement but about capability. Want to design, analyze, or fix code? AI can get you closer.
-
For managers: AI can reduce internal dependencies, but it also demands revisiting processes and responsibilities. Who should be in charge of critical tasks that someone else can now handle with AI?
-
For policy makers and trainers: the labor market can reorganize before statistics show it. Investing in practical training on AI tools is a bet to make transitions more equitable.
Final reflection
OpenAI’s research gives us an early window into how work is being reconfigured: not just automation or replacement, but a recomposition of tasks and roles. AI makes it easier for the person facing a problem to try to solve it in the moment, changing how we work and collaborate. Will you adapt or wait for the job description to change?
Original source
https://openai.com/index/how-ai-is-expanding-what-people-do-at-work