Your Job Description Is Already Out of Date
Job boundaries used to be legible. A salesperson sold, an analyst analyzed, a developer built. You could look at someone's title and guess, with decent accuracy, what was actually in their inbox. That's getting harder, and OpenAI's economic research team just put a number on why.
They looked at more than 800,000 work-related messages from U.S. ChatGPT users and asked something simple: when someone brings a work task to AI, does that task actually belong to their occupation? Their new report, Work at the Frontier: How AI Is Expanding What People Do at Work, says no, surprisingly often.
The headline figure: 16.8% of all work-related messages, and 43.5% of messages tied to a specific occupation, involve tasks historically owned by a different occupation. A small-business owner drafts contract language without calling a lawyer. A salesperson digs through a customer dataset that would've gone to an analyst a few years back. A marketer fixes a website bug instead of filing a ticket and waiting on a developer.
OpenAI calls this task crossover — work that's assigned to one occupation on paper, but keeps showing up in the AI usage of people who hold a different title.
Who's doing the most borrowing
Strip out the generic stuff — writing, summarizing, scheduling, the tasks every job touches regardless of role — and some occupations barely resemble their own job description anymore. Once you exclude that generic layer, here's the share of messages that fall outside a worker's own occupation:
- Customer experience workers: 77%
- Designers: 75%
- HR workers: 69%
- Legal workers: 56%
- Marketers: 53%
For several of these roles, most of the occupation-specific AI use isn't actually work from their own occupation. That's not a rounding error, that's most of the job.
Some tasks travel further than others
Crossover doesn't run the same direction for every field. OpenAI splits it into occupations that mostly borrow from others, and occupations that mostly supply work to others.
Design borrows hard: 35.2% of designers' messages pull in tasks from other fields, while design tasks make up only 1.7% of everyone else's messages. Designers constantly reach outside their lane. Almost nobody reaches into it.
Engineering is the mirror image. Just 18.5% of engineering messages involve outside tasks, but engineering work — troubleshooting, wrangling technical systems — shows up in 7.4% of messages from people in totally different roles. Engineering exports more than it imports.
Marketing does both, more than anyone. Marketers spend 24.3% of their messages on tasks that belong to other occupations, and marketing tasks account for 8.9% of everyone else's messages — the highest outward share in the whole dataset. Two tasks travel especially far across every occupation studied: financial calculation and technology troubleshooting, each landing in the top three outside-tasks for all seven other occupation groups.
Smaller teams absorb more of it
Company size shapes this too. In workspaces with 2–5 seats, 18.9% of the average user's messages fall outside their own occupation. In workspaces over 100 seats, that number drops to 16.3%. Makes sense — in a small team there's often no specialist to hand a problem to, so whoever hits it first just picks it up with AI's help. In a bigger company, there's usually still someone whose job that technically is.
That pattern flattens out for the heaviest AI users, where crossover looks similar no matter the company size. Maybe those people have settled into AI-assisted workflows that don't care much what the org chart says.
The job description hasn't caught up yet
Titles and formal job descriptions update slowly. Usage data doesn't wait around for that. What OpenAI is really showing is that AI adoption is an early signal — you can see roles reorganizing in the data long before it shows up in an org chart, a job posting, or a performance review template.
If a marketer is already doing a quarter of an analyst's job with AI's help, or a designer is writing copy a writer used to own, that's not a future-of-work thought experiment. It's already happening. Nobody's gotten around to writing it down.
