From Assistance to Execution: How Enterprises Put AI to Work

OpenAI has published a new research report on enterprise AI adoption, drawing on usage data from ChatGPT and Codex across its business customers. The headline finding: companies are moving from using AI to assist with analysis toward letting it execute work directly, and the gap between “frontier” firms and everyone else is widening fast.

Details

• The usage gap is tripling: Frontier firms — the top 10% of enterprise adopters — now generate 8.3x more output tokens per active user than typical firms, up sharply from just 2.6x in January
• Codex is overtaking ChatGPT inside enterprises: As of June, Codex accounts for 64% of combined Codex and ChatGPT output tokens among enterprise customers, reflecting a shift toward agentic coding work
• Advanced features separate the leaders: At frontier firms, 21% of weekly active users use Plugins (versus 9% at typical firms) and 19% use Skills (versus just 3%). Internally, OpenAI says 95% of its own employees use Plugins weekly
• Codex adoption is exploding outside engineering: Since February, Codex usage has grown 108x in legal, 41x in both sales and recruiting, 26x in marketing, and 5x in engineering — showing agentic coding tools spreading well beyond developer teams
• Junior employees message more than executives: Six months after adoption, early-career employees send 13 more messages per week than executives, suggesting adoption patterns skew toward people doing hands-on task execution
• Built on two underlying studies: The report synthesizes findings from “Enterprise Signals,” a look at agentic AI usage patterns, and “How Organizations Use AI: Evidence from ChatGPT,” which tracks adoption trajectories by role and seniority

What happened next

OpenAI frames this as a distinct, enterprise-focused companion to its earlier consumer-facing “Signals” report on global ChatGPT usage. Where that report tracked how individuals use ChatGPT day to day, this one is squarely about organizational adoption curves — and the message to enterprise customers is that the gap between companies that connect agents into real workflows with proper governance and those that treat AI as an individual productivity tool is compounding rather than closing. The full report and methodology are available on OpenAI’s blog at https://openai.com/index/how-enterprises-put-ai-to-work.