OpenAI reports Codex now dominates internal AI usage as agents shift work to long-horizon tasks

On June 25, 2026, OpenAI released an economic research paper measuring Codex's economic potential and its impact on knowledge work. The company states that agentic AI has shifted the unit of knowledge work from short, self-contained chatbot interactions to delegated, long-horizon tasks that agents can pursue independently for minutes or hours while orchestrating tools and iterating toward solutions.

According to the post, Codex has become the dominant AI tool inside OpenAI. Through August 2025, the average OpenAI worker spent less than 10% of their tokens on Codex; now, every department—including non-technical teams such as Legal, Finance, and Recruiting—uses Codex as its primary AI tool for work. For the average OpenAI worker, Codex usage accounts for more than 85% of output tokens, and across the company Codex represents 99.8% of weekly output tokens.

The paper documents four trends through May and June 2026. By May 2026, 80.6% of sampled individual users made at least one Codex request estimated to exceed 30 minutes of human work, 70.2% made one exceeding one hour, and 25.6% made one exceeding eight hours. Non-developer adoption grew especially rapidly, rising 137-fold among individual users and 189-fold among organizational users since August 2025; within OpenAI, non-developer usage rose 12-fold. By June 2026, the heaviest daily users at the 99th percentile regularly generated more than 60 hours of Codex agent turns per day across multiple parallel agents.

OpenAI also notes that Codex enables non-technical workers to perform tasks outside their typical job descriptions. While engineering and coding remain the largest work categories for technical departments, over one-fourth of Codex usage by workers in business functions involved engineering or coding tasks such as automation, data transformation, tooling, debugging, and structured analysis.

Source: OpenAI Blog