A quarter of individual users now ask OpenAI's Codex agents to complete tasks that would take humans more than eight hours, according to new economic research from the company.

The study, published Thursday, tracks how AI agents have shifted work patterns from brief chatbot interactions to long-horizon delegated tasks. By May 2026, 80.6% of sampled individual users made at least one Codex request estimated to exceed 30 minutes of human work, while 70.2% requested tasks exceeding one hour.

Within OpenAI itself, Codex now accounts for 99.8% of weekly output tokens generated by employees. Every department — including Legal, Finance, and Recruiting — crossed over to majority Codex usage by April 2026, with the average worker generating more than 85% of their tokens through the agent platform.

Non-developers driving fastest growth

Non-developer adoption has outpaced traditional coding use cases across all user segments. Individual non-developer users multiplied 137 times since August 2025, while organizational non-developer users increased 189-fold.

The heaviest internal users now generate more than 60 hours of Codex agent work per day through parallel task orchestration. Research teams saw the biggest usage jump, with median use rising 56 times higher than November 2025 levels.

More than one-fourth of work completed by OpenAI's business function employees through Codex involved engineering or coding tasks — work previously requiring technical specialists. The agent system enables workers to cross traditional skill boundaries and tackle adjacent responsibilities.

"Agentic AI changes the unit of knowledge work from single interactions to delegated, long-horizon tasks," the research paper states. "Agents can operate independently for minutes or hours while orchestrating tool calls, interacting with environments, and iterating towards solutions."

The data suggests agents are becoming the primary interface for AI-assisted work, replacing shorter chatbot exchanges with sustained autonomous execution across departments and skill levels.