OpenAI reports rapid growth in agentic AI use among EU business customers

The report measures use within OpenAI's customers

OpenAI's EU economic research announcement says Codex and ChatGPT Work account for more than half of output tokens in its EU enterprise sample. It also reports that weekly active users of those tools among EU small and medium businesses grew more than eightfold since May 2026.

The analysis defines smaller businesses as those with 200 or fewer associated users. These are provider-observed usage measures within a customer sample. They are not a census of European companies or a direct measure of work completed successfully.

Usage differs substantially across the sample

The reported share of enterprise output tokens attributed to Codex and ChatGPT Work ranges from 22% in Slovenia to 85% in Lithuania. OpenAI also defines frontier firms as the top 10% by output tokens per active user and reports a gap between those firms and typical customers.

Token volume can describe how intensively a tool is used. It cannot, by itself, establish that a worker saved time, that an agent completed a task correctly or that increased use caused a business improvement. Those questions require outcomes measured alongside the activity.

Choose a completion metric for a local trial

Franklin analysis: a business comparing its own adoption with this report should first define a useful unit of completed work. For a coding workflow, that might be a change accepted after review and testing. For a research task, it might be a source-supported answer that passes the team's checks.

Track elapsed time, corrections and reviewer effort as well as usage. Compare similar tasks before expanding access, and inspect unsuccessful attempts rather than counting them as productive activity. The report identifies a change in how customers use AI; an organization still needs its own evidence to decide whether deeper use improves the work it cares about.