What happened
A new OpenAI research paper shows that Codex, the company's agentic AI coding tool, has become the primary AI tool for every department at OpenAI, including non-technical ones like Legal and Recruiting. By May 2026, 80.6% of sampled individual users made at least one Codex request estimated to exceed 30 minutes of human work, and 25.6% made one estimated to exceed eight hours.
Why it matters
The paper documents a shift from short chatbot interactions to longer, delegated tasks that agents can handle independently. Codex usage now accounts for more than 85% of output tokens for the average OpenAI worker, and 99.8% of weekly output tokens generated within OpenAI, signaling a move away from chatbots as the primary AI interface for work.
What to watch
Non-developer adoption is outpacing developer adoption, with non-developer individual users rising 137x, organizational users 189x, and OpenAI users 12x since August 2025. The paper suggests this trend could redefine job roles, as over one-fourth of work done with Codex by workers in business functions was engineering or coding, potentially lowering the cost of moving across task boundaries.
Summaries like this, in your inbox every morning.
The paper's central claim is that the unit of knowledge work is changing from single interactions to delegated, long-horizon tasks. The evidence comes from OpenAI's own internal usage, where Codex adoption grew in tandem with its capabilities. As Codex leveraged stronger models and new features, it became capable of taking on more tasks, and usage patterns shifted accordingly.
The data shows a clear progression: engineers at OpenAI first shifted the majority of their usage to Codex by December 2025, followed by Legal, Finance, and Recruiting around April 2026. Notably, non-technical departments made the transition much faster once they started. This pattern suggests that as agentic tools become more capable, they move beyond their initial developer audience into broader knowledge work, allowing workers to do adjacent tasks that previously required specialized technical support.
The stakes for businesses and workers hinge on whether this internal OpenAI experience reflects the broader market trajectory. The paper itself notes this is "likely to be what the future of work looks like," but the findings are based on frontier users with broad, low-friction access to the tool. The real test will be whether the adoption trends observed inside OpenAI generalize to external individual and organizational users, especially as the tools continue to improve and handle longer, more complex tasks.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
Anthropic's internal study shows junior new hires in technical departments fell from 27 percent to 9 percent s…

Mastercard is testing a new intelligence layer in its Agent Pay Trust Framework in the United States that prod…

The iShares Semiconductor ETF, which holds leading AI chip suppliers, is in bear territory after declining mor…

In a decision dated September 24, 2026, the Aachen Regional Court banned Google from spreading three false AI…

Three designs for turning encoded inputs into predictions: pooling all inputs into one shared vector, applying…

Across 24 counted runs of /code-review on a 41-line shopping-cart file with six planted bugs, --max-findings 1…
