🎯 Impact Sentiment: Positive
📋 Summary
- OpenAI's economic research paper documents how agentic AI (specifically Codex) has become the primary work tool at OpenAI itself — accounting for 99.8% of weekly output tokens generated internally.
- By May 2026, 80.6% of individual Codex users made at least one request estimated to exceed 30 minutes of human work, with 25.6% making requests estimated at 8+ hours of human-equivalent output.
- Non-developer adoption grew faster than developer adoption: individual non-developer users increased 137x since August 2025, while OpenAI's own Legal, Finance, and Recruiting teams now rely on Codex as their primary AI tool.
- The shift marks a fundamental change in how knowledge work is structured — from single interactions to delegated, long-horizon tasks that agents execute independently.
💡 JR Insights
- 💼 Implication: When AI agents can perform hours of knowledge work autonomously across every department — including legal and recruiting — the scope of roles that can be reduced or augmented is vastly larger than most workers realize.
- 🚨 Risk: Roles defined primarily by executing well-defined, repeatable knowledge tasks (data analysis, document review, scheduling, standard coding) are the most immediately vulnerable to agentic AI replacement.
- ✨ Takeaway: Focus on task orchestration and judgment — learn to direct and validate AI agent outputs rather than competing with them on execution speed and volume.