How Agents Are Changing the Way We Work

The unit of knowledge work is shifting from one-off interactions to long-horizon tasks that can be delegated. While traditional chatbots provided short, independent answers, Agents operate independently for minutes to hours through tool calls, environment interaction, and iterative problem-solving.

According to OpenAI's internal data, Codex has already established itself as the core AI tool across every department. Starting with engineering, it has been adopted by non-technical departments including legal, finance, and recruiting, and currently more than 85% of output tokens generated by the average OpenAI employee come through Codex.

Key trends of change include the following:

• Long-horizon task execution: As of May 2026, 80.6% of individual users requested workloads of 30 minutes or more from Codex, and 25.6% requested workloads of 8 hours or more.
• Surge in non-developer adoption: Since August 2025, non-developer usage has grown 137x among individuals and 189x among organizations, outpacing the rate of developer adoption.
• Expanding scope of work: Non-technical users are also leveraging Codex to perform technical execution tasks such as automation, data transformation, and debugging.

Notably, the top 1% of heavy users are maximizing agent utilization to the extreme, running multiple agents in parallel and generating more than 60 hours of agent turns per day.