OpenAI Data Shows Enterprise AI Use Widening
Highest usage companies generated 8.3 times more AI output per user, while Codex use spread beyond engineering.
Topics
News
- Cisco Warns Critical Email Gateway Flaw Under Active Attack
- Bank of America Hires 1,000 Contractors as AI Raises Cyber Costs
- Netflix Cuts About 1,000 Jobs in Biggest Layoff Since 2022
- Indian Embassy in Jakarta Launches AI Avatar Dewi
- Anthropic Tests How Small AI Models Can Replicate Across IoT Devices
- Scientists Call for Global Oversight of Cyber-Capable AI
OpenAI’s heaviest enterprise users are pulling away fast from the rest, with the gap in AI output per user more than tripling in five months, according to new company data.
In June, the top 10% of OpenAI’s enterprise customers by usage generated 8.3 times more output tokens per active user than firms around the middle of the distribution. In January, the difference was 2.6 times.
OpenAI uses output tokens as an indicator of how extensively customers use its systems. The metric can capture longer, multi-step tasks performed by AI agents, but does not by itself measure productivity, cost savings or financial returns.
The findings come from two studies examining how businesses use ChatGPT and Codex, including an analysis of more than 10 million enterprise messages.
The data also points to AI use shifting from generating answers toward carrying out work. By June, Codex accounted for 64% of combined ChatGPT and Codex output tokens among enterprise customers, compared with 36% for ChatGPT.
Codex use is also expanding outside software development. Since February, weekly active users increased 108-fold in legal roles, 41-fold in sales, 41-fold in recruiting and 26-fold in marketing. Engineering users increased fivefold over the same period, although that function started from a more established base.
Heavy users were also more likely to connect AI systems with workplace tools and reusable instructions. Among employees at the highest-usage firms, 21% of weekly active users used Plugins, compared with 9% at typical firms. For reusable skills, the figures were 19% and 3%, respectively.
OpenAI’s data also found heavier usage among junior employees than senior executives. Six months after adoption, early-career workers sent around 13 more messages per week than executives.
The studies suggest that access to the same AI models is producing very different levels of adoption inside companies. They do not establish whether heavier usage translates into better business performance, leaving productivity and return on investment as the more important measures for enterprises to demonstrate.
