Agentic AI is shifting the unit of knowledge work from short chatbot interactions to delegated, long-horizon tasks that can operate independently for minutes or hours while orchestrating tool calls and iterating towards solutions. OpenAI‘s agentic tool Codex exemplifies this transformation. Within OpenAI, Codex has become the primary AI tool for every department, including non-technical ones like Legal, Finance, and Recruiting. The average OpenAI worker now generates more than 85% of their output tokens with Codex, and Codex accounts for 99.8% of weekly output tokens within the company.

Four adoption trends are documented. First, people use Codex for longer-horizon work. By May 2026, 80.6% of individual users made at least one Codex request estimated to exceed 30 minutes of human work, 70.2% exceeded one hour, and 25.6% exceeded eight hours. Second, Codex became the primary AI tool across all OpenAI departments: Engineering first, then Legal, Finance, and Recruiting by April 2026. Third, non-developer adoption outpaced developer adoption, with non-developer individual users rising 137x since August 2025, organizational users 189x, and OpenAI non-developers 12x. Fourth, Codex enabled OpenAI workers to do tasks outside their job description, such as non-technical employees using Codex for coding, automation, data transformation, and debugging.

Nearly a quarter of all Codex requests correspond to tasks that would take a person more than one hour. The heaviest daily active users at OpenAI generate over 60 hours of Codex agent turns per day across multiple parallel agents. As Codex became more powerful and parallelizable, users moved from single-answer interactions to orchestrating multiple agent tasks over a day. Median internal usage jumped 56x in Research, 32x in Customer Support, 27x in Engineering, and 13x in Legal since November 2025.

Non-developers are the fastest-growing user group. While Codex began as a coding tool for developers, adoption among non-developers has grown even more quickly. This does not mean every non-developer uses Codex like an engineer, but that more non-developers are using Codex for some kind of agentic work. A heat map of occupation versus work done with Codex shows that engineering and coding are the largest category for data science and research, while knowledge work is largest for finance, business operations, marketing, and other departments. Notably, over one-fourth of work done with Codex by workers in business functions was engineering or coding, indicating that agentic tools lower the cost of crossing task boundaries.

The economic potential is significant. Increased use of agentic tools by non-engineer employees expands the frontier of what workers can do, affecting workflow redesign, skill valuation, and labor market policy. The paper’s findings demonstrate that when people have broad, low-friction access to capable agentic tools, they use them for longer, more complex, and more cross-functional work, suggesting this will be the future of work.