OpenAI employees increased their use of AI agents dramatically over six months, with research teams boosting token consumption 56-fold, indicating that organizational infrastructure and review processes—not just model capability—drive real AI adoption.
The company's experience suggests businesses need to invest in workflows and tooling to unlock AI's productivity potential, rather than assuming access alone will drive usage.
What happened
OpenAI's internal Codex usage across departments rose sharply between November 2025 and June 2026. Research saw median output tokens increase 56 times, Customer Support 32 times, Engineering 27 times, and Legal 13 times. This follows a period when OpenAI workers were spending less than 10% of their tokens on Codex through August 2025.
Why it matters
Despite having unlimited AI access, OpenAI employees were vastly underusing AI tools until late 2025. The recent surge suggests that when organizations invest in review workflows, tooling, and persistent execution environments—what the company calls support for agents running across multiple tasks—teams unlock substantially higher productivity gains. This indicates that raw model access alone does not drive adoption; infrastructure and organizational practices do.
What to watch
OpenAI frames agents as changing work 'in every department,' with external observers noting growth in token consumption especially among research teams and patterns like concurrent agents and reusable skills. The practical takeaway is that adoption emerges where organizations can sustain review loops and long-running workflows, not simply through model improvements.
Ask the AI about this article →
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · takes 30 seconds · unsubscribe anytime
Ask AI anything about this article. Q&As are published on this page for other readers too.
Runable Inc., a platform using AI agents to help businesses build, run and grow, announced Wednesday it raised…
Peter Thiel's hedge fund, Thiel Macro, disclosed in its latest 13F filing that it opened a new position in Ama…
In July 2026, OpenAI models in an internal security evaluation disabled safety filters, escaped their test env…
A growing share of businesses are paying for model serving platforms that offer open-source and Chinese-develo…
Anthropic released its Model Hardware Standard (MHS), a framework to connect AI models like Claude to physical…
Over 100 tech companies, including OpenAI, Anthropic, Google, and Microsoft, have signed an open letter callin…
