OpenAI: Frontier firms generate 8.3× more AI output tokens per active user
OpenAI's enterprise research shows a widening gap in AI usage depth: frontier firms—the top 10% of adopters—now generate 8.3 times more output tokens per active user than typical firms, up from 2.6 times in January. The shift reflects a move from answering questions to carrying out work, with agentic tools like Codex producing most output. Notably, frontier firms adopt advanced capabilities like Plugins (21% vs 9%) and skills (19% vs 3%) much more heavily, though even their usage trails OpenAI's internal 95% Plugin adoption. The reports suggest access alone isn't enough; complementary investments in workflows, data, and governance matter. Caveat: the data stems from OpenAI's own customer base, so selection bias may inflate the divergence.
OpenAI: Frontier firms generate 8.3× more AI output tokens per active user
OpenAI: "As of June, frontier firms generated 8.3× as many output tokens per active user as typical firms, a threefold increase compared to the 2.6× gap in January. Read together, the reports suggest that access alone may not be enough to scale AI."
Key takeaway
Execution depth, measured by tool and plugin usage, is the new metric for enterprise AI maturity; the gap between frontier and typical firms tripled in five months.
What happened
OpenAI's August 12 research reports that frontier firms—the top decile of enterprise customers by monthly output tokens per active user—now generate 8.3 times more output tokens per active user than typical firms, up from a 2.6 times gap in January, a threefold increase.
The shift is driven by adoption of advanced capabilities: among weekly active users, 21% at frontier firms use Plugins and 19% use skills, versus 9% and 3% at typical firms. Codex generated 64% of combined Codex and ChatGPT output tokens among enterprise customers as of June, and Codex usage grew 108× in legal since February.
Evidence
As of June, frontier firms generated 8.3× as many output tokens per active user as typical firms, up from a 2.6× gap in January.
OpenAI News · attributed
"As of June, frontier firms generated 8.3× as many output tokens per active user as typical firms, a threefold increase compared to the 2.6× gap in January."
Among weekly active users, 21% at frontier firms use Plugins and 19% use skills, compared with 9% and 3% at typical firms.
OpenAI News · attributed
Among weekly active users, 21% at frontier firms use Plugins and 19% use skills, compared with just 9% and 3% at typical firms.
As of June, Codex generated 64% of combined Codex and ChatGPT output tokens among enterprise customers.
OpenAI News · attributed
As of June, Codex generated 64% of combined Codex and ChatGPT output tokens among enterprise customers.
Since February, weekly active enterprise Codex users grew 108× in legal, 41× in sales, 41× in recruiting, and 26× in marketing, compared with 5× in engineering.
OpenAI News · attributed
Since February, weekly active enterprise Codex users grew 108× in legal, 41× in sales, 41× in recruiting, and 26× in marketing, compared with 5× in engineering.
Why it matters
This widening token-per-user gap signals that access alone won't deliver value; builders must invest in integrations, skills, and governance to close the adoption cliff and convert assistance into execution.
Limits and uncertainties
The data is drawn from OpenAI's own enterprise customer base, which may not represent the broader market.
Output tokens are a proxy for depth of use and may not capture quality or business outcomes, and the reports are published by OpenAI, introducing potential selection bias.
Practical implications
Operators should build robust plugin ecosystems and skill libraries, as these are the primary differentiators for high-intensity enterprise adoption.
Invest in governance and permission frameworks early to support agentic workflows, and help employees turn individual workflows into shared ways of working to accelerate adoption.
What to watch
Whether typical firms close the gap in plugin/skill adoption, or if the frontier continues to pull ahead.
If Codex usage spreads beyond engineering into legal and sales at the observed growth rates.
If OpenAI's internal ~95% plugin usage becomes a benchmark for enterprise deep adoption.
