OpenAI Says Most EU Business AI Is Agentic: A Number Worth Examining

🔍 Read the full analysis: OpenAI Says Most EU Business AI Is Agentic: A Number Worth Examining on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get hardware and tech essentials delivered free with Prime

• Fast, free delivery on millions of items
• Prime Video, Amazon Music and more included
• Member-only deals all year

Start your free Prime trial →Free trial for eligible customers · Cancel anytime

As an affiliate, we earn on qualifying purchases.Hide

OpenAI says its AI agents generated 55% of output tokens across its ChatGPT Enterprise and ChatGPT Business accounts in the European Union by September 30, a finding that points to growing agent use among its customers but does not show that agents handle most business AI tasks. The company published the figure on October 14 in a report covering accounts in the 27 EU member states.

OpenAI Economic Research’s report, Europe’s Agentic Opportunity: Moving from breadth to depth with AI agents, groups Codex and ChatGPT Work under the term “agentic”. It says Codex accounted for 34% of output tokens and ChatGPT Work for 21% by September 30. The remaining 45% came from standard ChatGPT use. The report covers the company’s Enterprise and Business accounts, not every European organization using AI.

OpenAI also reports rapid growth in agent use among businesses with 200 or fewer users: weekly active users of Codex or ChatGPT Work rose 8.1 times from the first week of May to the week ending October 5. The company says every member state grew at least fivefold, with Belgium at 12.1 times, Italy at 10.8 times and Luxembourg at 10.0 times. It does not provide absolute user counts for that measure.

The report’s country figures vary widely. Agentic tools accounted for 85% of output tokens in Lithuania and 81% in Cyprus, compared with 22% in Slovenia and 26% in Austria. Germany was at 51%, ranking 17th among the 27 member states. These are shares within OpenAI’s measured accounts, not national estimates of AI use.

What the 55% Figure Measures

The report offers a view of how OpenAI’s business customers are using its newer tools, and the increase suggests that agents are taking a larger role in the company’s products. But output tokens are not tasks, users or economic value. An agent that reads files, plans work, produces a draft and checks its result may generate many more tokens than a short chat response. A larger token share can reflect that heavier processing as well as broader adoption.

OpenAI itself cautions against treating tokens as a direct measure of business value. Its Enterprise Signals material describes tokens as an imperfect proxy and says agentic tasks can require substantially more computation. The 55% figure is therefore best read narrowly: agents produced more of the text generated in these accounts than standard ChatGPT did. It does not establish that agents carried out most AI interactions, improved productivity, or delivered a particular financial return.

The distinction matters for companies and policymakers weighing how fast AI work is changing. A product-use measure can help identify where adoption is occurring, but decisions about skills, investment and productivity require evidence about actual work completed, quality, cost and outcomes. The report does not provide those measures.

A New Tool Shapes the Trend

The growth period includes the launch of ChatGPT Work, which OpenAI says became available in July 2026. The tool can work with files, browsers and company tools. Codex, the company’s coding agent, was already contributing output in May. In OpenAI’s chart, Codex represented roughly a fifth of EU output tokens in May and 34% by September 30; ChatGPT Work contributed 21% by the September date.

That timing affects how the growth should be understood. OpenAI says a major part of the increase in small-business weekly active users followed the launch of ChatGPT Work. The 8.1-times comparison starts in the first week of May, before the product existed. It records growth in use of the two products over that period, but cannot by itself separate the effect of the launch from other changes in customer behavior or OpenAI’s customer base.

The report also compares high-use companies. It defines “frontier firms” as the top 10% of firms by output tokens per active user. In September, EU frontier firms generated about four times as many tokens per user as typical EU firms, while frontier firms outside the EU generated more than twice as many as the EU frontier group. OpenAI says EU companies account for about 10% of frontier firms worldwide, compared with about 65% in North America. These comparisons use the report’s own customer and usage definitions.

What the Data Cannot Show

OpenAI has not disclosed account or user sample sizes for each country in the material summarized here, making it difficult to judge how representative the national percentages are. The company notes that smaller countries, including Cyprus, Estonia and Lithuania, have relatively high agentic use within their larger firms. A small number of large customers could affect a country’s figure; the report does not provide enough detail to quantify that possibility.

The report also provides no absolute weekly active-user counts behind the 8.1-times small-business increase. Without the starting and ending totals, readers cannot tell how large the measured user group was. Nor does the token share establish how many business tasks agents completed, whether those tasks were successful, or what value customers received. OpenAI’s findings describe its own accounts, and the source does not establish how closely those customers represent EU businesses overall.

From Adoption to Business Outcomes

OpenAI’s report calls for European businesses to move from broad experimentation toward deeper use, with more worker skills and confidence and wider access to advanced tools. The company presents the data as evidence for that policy direction; it is not an independent assessment of the EU economy.

The next useful evidence would include absolute user counts, country-level sample sizes, task-level usage and measures of cost, quality and business outcomes. Until such data is available, the report provides a snapshot of agent use in OpenAI’s EU business accounts—not a complete measure of how agentic European business AI has become.

Key Questions

What does OpenAI mean by “55% of output tokens”?

It means Codex and ChatGPT Work generated 55% of the output tokens from the measured ChatGPT Enterprise and Business accounts in the EU by September 30, 2026. It is a share of generated text, not a share of users, tasks or business value.

Does the report show that most European businesses use AI agents?

No. The data covers OpenAI’s business accounts in the EU, not all European companies. It does not establish the proportion of businesses using agents across the region.

What drove the reported 8.1-times growth in small-business use?

OpenAI measured weekly active users of Codex or ChatGPT Work at businesses with 200 or fewer users, comparing the first week of May with the week ending October 5. The company says a major part of the increase followed ChatGPT Work’s July launch. It did not provide absolute user totals.

Can the country rankings show which EU nations are most advanced in AI?

Not on their own. The percentages describe agentic output within OpenAI’s accounts in each country, and OpenAI did not provide country-level account or user counts in the source material. They should not be treated as rankings of national AI readiness.

Source: ThorstenMeyerAI.com