How workers are unlocking new ways of working

OpenAI’s Global Affairs post says new Economic Research shows workers using AI for tasks historically tied to other occupations—and returning to those tasks until they become a larger share of observed AI use. The study analyzes more than 1.5 million work-related ChatGPT messages from April through July 2026.

Titles stay put while task bundles widen. That’s a quieter labor story than mass layoff headlines—and potentially a bigger one for how jobs actually change.

Companies guessing at AI ROI need this kind of task-level evidence, not only seat licenses.

We’re reading this as job redesign from below.

OpenAI’s earlier Work at the Frontier report flagged task crossover; this follow-up asks what sticks. Recurring cross-occupation use that grows as a share of a worker’s AI activity suggests experimentation becoming habit. That’s useful AI: a marketer who can draft SQL once, an engineer who can structure a brief—friction between noticing a problem and acting drops.

The caveats are load-bearing. This is OpenAI studying ChatGPT messages—vendor telemetry, not a random labor survey. “Work-related” classification and occupation labeling can misread context. And broadening tasks without rewriting job descriptions can mean unpaid scope creep.

I’m watching whether employers formally recognize crossover skills—or quietly expect them without pay. Also whether non-ChatGPT tools show the same pattern in independent studies.

OpenAI Economic Research previously published Work at the Frontier on task crossover. The new analysis covers April–July 2026 professional ChatGPT use.