OpenAI says 98% of staff use Codex agents, but data is self-reported
▼ Summary
– OpenAI reports that 98% of its employees now use Codex, up from 40% in August 2025, but all metrics are self-reported by the company that sells the product.
– Non-developer usage of Codex grew 137 times since August 2025, and non-developers now make up roughly 20% of the platform’s five million weekly users.
– OpenAI’s legal team generated 13 times more tokens in June than in November 2025, which the company presents as evidence of agent adoption beyond engineering.
– The paper does not address whether OpenAI incentivizes Codex use or present data on whether the shift measurably improved output quality or reduced completion time.
– The self-reported data serves as a signal for investors before a potential Q4 IPO, but independent measurement is needed to verify the claimed pace and scale of adoption.
Nearly 98 percent of OpenAI’s workforce now relies on Codex, the company’s AI coding agent, according to a paper released Wednesday titled “The Shift to Agentic AI: Evidence from Codex.” That figure marks a dramatic jump from roughly 40 percent in August 2025. The report describes a sweeping transformation inside the company, where employees have moved from conversational chatbot interactions to deploying autonomous agents that handle multi-step tasks. However, every metric cited in the paper is self-reported by OpenAI, a company with a clear financial stake in promoting the product it is measuring.
The numbers are undeniably eye-catching. Active Codex users surged fivefold during the first half of 2026, while requests for tasks estimated to require eight or more hours of work jumped nearly tenfold. OpenAI’s legal team alone generated 13 times more tokens in June than in November 2025, a figure the company uses to argue that agents are now penetrating departments far removed from engineering.
OpenAI’s narrative centers on non-developer adoption, which has skyrocketed. Individual non-developer usage of Codex grew 137 times since August 2025, organizational non-developer usage rose 189 times, and internal non-developer adoption expanded twelvefold. The company expanded Codex earlier this month with enterprise plugins that connect to 62 business applications. Non-developers now account for roughly 20 percent of the platform’s five million weekly users, adopting the tool three times faster than engineers.
The paper frames these trends as evidence of a broader market shift from chatbots to agents. OpenAI argues that the pattern visible inside its own walls, where every department from legal to recruiting treats Codex as a primary tool, previews the trajectory of enterprise AI adoption. The company points to external data showing Codex usage among organizations at roughly 17 percent and among individuals at under one percent, suggesting substantial room for growth.
Yet the gap between internal and external adoption raises questions about how representative OpenAI’s workforce actually is. The paper does not address whether the company incentivizes or encourages employees to use Codex, a relevant omission given that near-universal adoption inside a company selling the product does not equal organic demand. No independent third party has verified any of the usage figures.
The self-reporting issue extends to productivity claims. OpenAI suggests that longer task requests and higher token generation prove agents are handling more complex work. As The Register noted, faster code generation does not automatically translate into proportional productivity gains, because verification, testing, and deployment time may expand to absorb the speed improvement. The paper offers no data on whether the shift to agents has measurably improved output quality or reduced total time to completion.
The broader context is a race among AI companies to prove that agents, not chatbots, represent the next phase of the market. OpenAI merged ChatGPT and Codex under Greg Brockman in May, consolidating its product strategy around a single agentic platform ahead of a potential Q4 IPO. Anthropic’s Claude Code and Google’s Gemini are pursuing similar strategies, making the competitive pressure to show adoption growth intense.
Meta’s internal experience offers a parallel data point. The company introduced a “Claudeonomics” leaderboard in April that tracks token consumption by team, turning AI usage into a visible performance metric. That approach, like OpenAI’s paper, measures input volume rather than output value, a distinction that matters when the companies reporting the numbers are also the ones selling the tools.
This paper is most useful as a signal of where OpenAI believes the market is heading, and where it wants investors to believe the market is heading before its IPO. The shift from chatbot queries to autonomous agent tasks is real and visible across the industry. Whether it is happening at the pace and scale that OpenAI’s self-reported data suggests is a question that only independent measurement will answer.
(Source: The Next Web)
