How agents are transforming work (openai.com)

A transformative shift in workplace AI adoption has been observed, with agentic AI tools like Codex significantly altering the nature of knowledge work. Unlike traditional chatbots, which handle short, isolated queries, Codex can autonomously perform complex, multi-step tasks over extended periods. This capability has dramatically increased its utilization across various departments at OpenAI, with Codex now responsible for over 85% of output tokens among workers, a stark rise from less than 10% in early adoption phases. By mid-2026, a significant number of users were leveraging Codex for tasks that would typically take several hours, showcasing a trend towards longer, more intricate projects. This evolution is especially notable as non-developer utilization skyrockets; since 2025, non-developer users have surged 137 times among individuals and 189 times within organizations. This rapid adoption across technical and non-technical fields hints at a future where AI tools are integral to transforming workflows beyond traditional boundaries, thereby enhancing productivity and reshaping job roles. The implications are profound, as organizations reconsider operational structures and workforce training, while policymakers evaluate broader impacts on labor markets driven by these advancements in AI capabilities.