← Back to feed

OpenAI published a research report on July 28, 2026 titled “Work at the Frontier: How AI is Expanding What People Do at Work.” The central finding is a challenge to the dominant job-displacement framing: AI is not primarily replacing what workers do — it is expanding the range of tasks they attempt.

The report draws on usage data from ChatGPT and a parallel study by Yang et al. examining query patterns on Perplexity. In both datasets, users regularly seek help with tasks that fall outside their stated primary occupation. An accountant asks for legal interpretation. A software engineer asks for copywriting. A doctor asks for financial modelling. The AI fills the cross-occupation gap rather than replacing the person’s core job.

The Framing Shift

The standard model in AI-and-jobs analysis assumes a fixed task bundle: identify which tasks an occupation involves, determine which ones AI can perform, and estimate displacement. The OpenAI report argues this framework breaks down when AI changes the cost of attempting new tasks. If trying something outside your expertise was previously prohibitive — because it required hiring someone or spending weeks learning — and AI makes it accessible in minutes, the relevant effect is scope expansion, not substitution.

This is empirically distinct from the claims being debated in most policy contexts. The AI-kills-jobs narrative concerns AI performing existing tasks at lower cost. The AI-expands-scope finding concerns AI enabling workers to attempt tasks they previously would not have. Both can be true simultaneously, but the policy and management implications are different.

What the Data Shows

The report does not publish raw displacement numbers — it is framed as a signal study, not a count of jobs lost or gained. The observable evidence is query patterns: users consistently requesting help with tasks labelled in occupational frameworks as belonging to other job categories. The Yang et al. finding from Perplexity data provides independent corroboration of the same pattern on a different platform.

The implication is that firm-level job descriptions — and the occupational frameworks used to model AI impact — may be behind the actual behavior of workers with access to frontier AI tools. Workers are already recombining task bundles that the org chart has not caught up with.

What It Does Not Resolve

The report is explicit that AI may change both how work gets done and who does it. Scope expansion does not rule out net job reduction. If one worker can now handle tasks that previously required two people — because AI lets them cover ground they could not before — headcount still falls. The report does not address that second-order effect directly.

The dominant counterclaim — that AI is already showing up in job cut announcements and hiring freezes — is also well-documented. AI has been the leading stated reason for US job cuts for two consecutive months. Those two facts are not in contradiction with the scope-expansion finding; they describe different dynamics running in parallel.

The Provenance

The report is authored by Caroline Chin and Alex Martin Richmond at OpenAI and published under the OpenAI Research imprint. OpenAI has an obvious institutional interest in a framing that positions ChatGPT as a capability amplifier rather than a job eliminator. The underlying query data is proprietary and not independently verifiable. The academic parallel in the Yang et al. Perplexity study adds external support, but the core dataset belongs to OpenAI.