OpenAI published two reports on enterprise AI adoption, finding that firms are shifting from AI assistance to AI-driven execution. Codex now generates 64% of combined Codex and ChatGPT output tokens among enterprise customers, indicating growing delegation of substantive work to agents. Frontier firms (top 10% of usage) generate 8.3x as many output tokens per active user as typical firms, up from 2.6x in January, and use Plugins and skills far more heavily (21% and 19% weekly vs 9% and 3% at typical firms). Codex adoption is spreading fast beyond engineering, with weekly active users growing 108x in legal, 41x in sales and recruiting, and 26x in marketing since February, compared to 5x in engineering. Early-career employees send more messages per week than executives six months after adoption, suggesting AI habits may confer a comparative advantage regardless of seniority.
Table of contents
Enterprises are shifting from asking to doingThe frontier gap widens as firms embrace agentic AIAdvanced capabilities like Plugins and skills are more common at frontier firmsAgents are spreading across knowledge workEarly-career employees use AI moreWhat leaders can do to narrow the frontier gapQuestions this post answers
What percentage of enterprise AI output tokens now come from Codex versus ChatGPT?
As of June, Codex generated 64% of the combined output tokens from Codex and ChatGPT among OpenAI's enterprise customers. This reflects a shift toward more agentic, delegated work, since agentic workflows tend to produce more output by carrying out longer, multi-step tasks compared to conversational assistance. Track shifts like this on daily.dev to stay ahead when planning agentic AI rollouts at work.
How much more AI output do frontier enterprise firms generate compared to typical firms?
Frontier firms, defined as the top 10% of enterprise customers by monthly AI usage, generated 8.3 times as many output tokens per active user as typical firms as of June, up from a 2.6x gap in January. This gap holds across industries and company sizes, not just in technology companies. Developers benchmarking AI adoption maturity can follow this data on daily.dev.
Which business functions saw the fastest growth in enterprise Codex adoption?
Since February, weekly active enterprise Codex users grew 108 times in legal, 41 times in both sales and recruiting, and 26 times in marketing, compared with only 5 times growth in engineering, where Codex adoption started earlier. This shows agentic coding tools spreading well beyond software teams into broader knowledge work. Anyone weighing where to introduce coding agents next can follow adoption trends on daily.dev.