OpenAI's agent adoption research shows work moving from short AI interactions toward delegated execution. The unit is no longer a prompt and a response. It is a task that can run for minutes or hours, use tools, inspect environments, and return with a result.
The strongest signal is internal adoption. OpenAI says every department, including Legal and Recruiting, now uses Codex as its primary AI tool for work.
Longer tasks change the adoption curve
OpenAI reports that by May 2026, 80.6% of sampled individual users made at least one Codex request estimated to exceed 30 minutes of human work. More than 70% made one estimated to exceed one hour.
That changes how teams should think about AI rollout. Adoption is not only about encouraging usage. It is about deciding which work can be delegated, how the result gets reviewed, and where the handoff lands when the agent finishes.
The boost spreads beyond technical teams
OpenAI says non-developer adoption grew especially fast, and that non-technical users regularly use Codex for automation, data transformation, tooling, debugging, and structured analysis.
That is the real limit shift. Agents let individuals cross boundaries that used to require another team, another ticket, or another specialist. The gain is not that everyone becomes an engineer. The gain is that more people can move blocked work forward without waiting for every small technical step to be handed off.
Related services: Automation, Enablement, Agent implementation