OpenAI published an October 14, 2026 economic research update about agentic AI use among businesses in the European Union. In its EU sample of ChatGPT Enterprise customers, Codex and ChatGPT Work now account for more than half of output tokens. This is a share of generated tokens in the sampled accounts, not a claim that agents perform more than half of all business tasks worldwide.
The company also reports that weekly active users of Codex and ChatGPT Work among EU small and medium businesses grew more than eightfold since May 2026. It defines those businesses as having 200 or fewer associated users. User growth is not the same as revenue growth, completed work, paid seats or a measured productivity improvement.
Adoption intensity varies considerably across countries in the sample. OpenAI says the share of output tokens generated through Codex and ChatGPT Work ranges from 22 percent in Slovenia to 85 percent in Lithuania. These figures describe customer-account usage in this analysis, not national adoption rates for every business or resident in those countries.
The update defines frontier firms as the top 10 percent of firms by output tokens per active user, using ChatGPT Enterprise and Business accounts. In September, frontier firms in the EU generated more than four times as many output tokens per active user as typical EU firms. Firms outside the EU at the global frontier generated more than twice the EU frontier level, according to OpenAI.
Output tokens measure generated model text and related output, not the quality or economic value of the work. Longer responses and different workflows can change token totals. The release does not establish that higher token use causes better business performance, or that one country's firms get a better return on their software spending than another's.
OpenAI presents these findings as evidence of a shift toward tools that carry out tasks and argues for deeper use, training and wider access. This is a vendor's analysis of its own customer usage, not a new product launch or an independent productivity trial. It provides no basis for the draft's comparison with Apple's model pricing, and does not announce a new ChatGPT plan or price.