OpenAI said its Codex tool has shifted from a developer-focused coding assistant into a broader work platform, with non-technical employees adopting it at a much faster pace in recent months. In a research post released Wednesday, the company said Codex is now the primary AI tool for every department inside OpenAI, including Legal, Finance and Recruiting.
The company said the change reflects a wider movement toward agentic AI, where systems do not just answer prompts but can carry out longer tasks, use tools and iterate over time. According to OpenAI, that capability is helping Codex take on work that previously would have been handled through shorter chatbot exchanges.
OpenAI said the transformation has been especially visible inside its own workforce. Through August 2025, employees were still spending less than 10% of their AI tokens on Codex on average, with ChatGPT remaining the default tool for work. By June 2026, the company said Codex had become the main AI system used across the organization.
A central finding in the report is that non-developer adoption has accelerated sharply. OpenAI said the number of non-developer users increased faster than developer users across its individual, organizational and internal populations. Since August 2025, non-developer usage rose 137 times among individual users, 189 times among organizational users and 12 times within OpenAI.
The company said this does not mean all non-technical employees use the software like engineers. Instead, it said more of them are turning to Codex for agentic work, including automation, data transformation, debugging, tooling and structured analysis.
OpenAI also said users are increasingly asking Codex to handle longer, more complex assignments. By May 2026, 80.6% of sampled individual users had submitted at least one request the company estimated would take more than 30 minutes of human time. The same was true for 70.2% of users for tasks estimated to take more than an hour, while 25.6% had made at least one request estimated to exceed eight hours.
The report said nearly a quarter of all Codex requests are now for work that would take a person more than one hour to complete.
OpenAI said engineering teams adopted Codex first, but other departments followed. It said Legal, Finance and Recruiting crossed into majority Codex usage around April 2026. The average engineer at the company now generates 99% of output tokens with Codex rather than ChatGPT, according to the report.
The company said the change has been accompanied by heavier use throughout the organization. By June 2026, users in the 99th percentile were regularly generating more than 60 hours of Codex agent turns in a single day, spread across multiple parallel agents.
OpenAI said different departments have adopted the tool at different speeds, with Research showing the strongest increase over the past six months. The median Research user’s combined output tokens were 56 times higher in June 2026 than in November 2025, according to the company. Customer Support was up 32 times, Engineering 27 times and Legal 13 times.
OpenAI said the findings matter for businesses redesigning workflows, workers deciding which skills will be most valuable and policymakers studying AI’s effects on labor markets. The company argued that greater access to capable agentic systems can expand the range of tasks employees can perform without specialized technical support.
In one of its examples, OpenAI said more than a quarter of work done with Codex by workers in business functions involved engineering or coding tasks. It said that suggests agent tools can lower the barriers between job categories and allow employees to move into adjacent work.
The company framed the report as evidence of what it called the economic potential of frontier agentic AI. Its conclusion was that as these tools improve and become easier to use, people are likely to rely on them for longer, more complex and more cross-functional work.