How To Read OpenAI’s Claim That Most EU Business AI Is Agentic

🔍 Read the full analysis: How To Read OpenAI’s Claim That Most EU Business AI Is Agentic on ThorstenMeyerAI.com

Age 18–24?Offer from Amazon

Prime made for students and young adults

• Fast, free delivery for dorm and study essentials
• Prime Video and Amazon Music included
• Member-only deals

Try Prime for Young Adults →Free trial for eligible 18–24 year olds

As an affiliate, we earn on qualifying purchases.Hide

OpenAI said its agent products generated 55% of output tokens in ChatGPT Enterprise and Business accounts across the 27 EU member states by Sept. 30, 2026. The figure, published in a report on Oct. 14, counts text generated by Codex and ChatGPT Work; it does not show that agents handled most business AI tasks or delivered most of the value.

OpenAI Economic Research’s report, Europe’s Agentic Opportunity: Moving from Breadth to Depth with AI Agents, defines “agentic” use as activity through Codex, OpenAI’s coding agent, and ChatGPT Work, an agent mode launched in July 2026 that can work with files, browsers and company tools. By Sept. 30, Codex accounted for 34% of output tokens and ChatGPT Work for 21%; plain ChatGPT accounted for the remaining 45%.

OpenAI also reported that weekly active users of Codex or ChatGPT Work at businesses with 200 or fewer users increased 8.1 times between the first week of May and the week ending Oct. 5. The company said every EU member state recorded at least fivefold growth. Belgium, Italy and Luxembourg had the largest reported multiples, at 12.1, 10.8 and 10.0, respectively. OpenAI did not provide absolute user counts in the source material.

The report’s country figures vary considerably: agentic products represented 85% of output tokens in Lithuania and 81% in Cyprus, compared with 22% in Slovenia and 26% in Austria. Germany’s reported share was 51%. OpenAI also compared its highest-use companies, defined as the top 10% by output tokens per active user, with other firms. It said 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.

What the Token Share Measures

The 55% figure describes where generated text went within a particular vendor’s business accounts. It is relevant as a sign that customers are using OpenAI’s newer agent products, but it is not a direct measure of how many employees use agents, how many tasks they complete, or whether those tasks improve productivity. Readers should not treat it as a measure of agent adoption across all European businesses.

Tokens are units of text processed or produced by a model. An agent may read files, plan a sequence of actions, generate code or documents, check results and try again. Such a workflow can use substantially more tokens than a short chat response. OpenAI’s Enterprise Signals page itself describes tokens as an imperfect measure of business value and notes that agentic tasks can require more computation. A larger share of tokens can therefore reflect longer, more involved runs, not necessarily a majority of interactions.

That distinction matters for business and policy decisions. The report argues that Europe should turn broad AI adoption into deeper use by building worker skills and confidence and making advanced tools more accessible and affordable. Those are the company’s policy conclusions, not outcomes established by the token figures. The reported gap between EU and non-EU frontier firms may point to a usage difference among OpenAI customers, but the data alone do not identify its causes or show what would close it.

Popular right now

KitchenAid 3-Piece Pasta Roller and Cutter Set for Stand Mixers

AmazonView Latest Price

eufy E25 Robot Vacuum and Mop with HydroJet Roller Mop, 20,000 Pa Suction

AmazonView Latest Price

Ninja Flip Toaster Oven and Air Fryer SP151

AmazonView Latest Price

As an affiliate, we earn on qualifying purchases.

A Launch Shapes the Growth Story

The reported rise includes a new product launched during the measurement period. Codex accounted for roughly a fifth of EU output tokens in May and 34% by Sept. 30, according to the report’s chart. ChatGPT Work launched in July and reached 21% by the end of September. That means the increase in agentic share cannot be read simply as existing users gradually changing how they work: a newly available product contributed substantially to the total.

The same timing affects the small-business growth statistic. Its starting point is the first week of May, before ChatGPT Work existed, and OpenAI says a major part of the increase followed the product’s introduction. An 8.1-fold rise from that baseline indicates increased active use of the included tools among the measured accounts. Without absolute counts, it does not reveal how many businesses or users were involved, or how representative those accounts are of small businesses throughout the EU.

OpenAI’s report covers only ChatGPT Enterprise and Business accounts in the 27 EU countries. It is not a survey of every AI provider or a census of European workplaces. Its figures should be understood as a view of activity on OpenAI’s own services, which the company also sells. That scope does not invalidate the measurements, but it limits what they can establish about the wider market.

What the Account Data Cannot Show

OpenAI has not supplied country-by-country account or user counts in the source material. The national percentages may be affected by which businesses use its products in each country, particularly where the customer base is smaller. The country rankings therefore describe the measured OpenAI accounts; they are not a reliable league table of national AI readiness.

The data also do not establish how many distinct tasks were agentic, how often agents completed work successfully, or whether their use produced better business results. The 8.1-fold statistic has no absolute user total in the report as described here, and its baseline predates ChatGPT Work’s launch. The report’s token figures cannot resolve those gaps.

OpenAI says EU frontier firms produce about four times the tokens per active user of typical EU firms, and that frontier firms outside the EU produce more than twice the EU frontier level. Those comparisons are based on the company’s own customer activity and its definition of frontier firms as the top 10% by output tokens per active user. The source material does not establish why the difference exists, whether it holds across other AI services, or how it relates to business performance.

Evidence Needed Beyond Token Counts

OpenAI’s report points toward a policy and business agenda: expand practical use, improve workers’ skills and confidence, and make advanced capabilities more accessible and affordable. The next useful evidence would show absolute user and account totals, task-level use and outcomes over a longer period, including comparisons that account for the July launch of ChatGPT Work.

For readers tracking the claim, the key question is whether future data show sustained agent use and measurable results, rather than a rising share of model output alone. The report does not provide a schedule for a follow-up release or independent verification of its account-level findings. Until such information is available, the figures are best read as a snapshot of OpenAI’s EU business customers through September and early October 2026.

Key Questions

What does OpenAI mean by “55% agentic”?

It means Codex and ChatGPT Work generated 55% of output tokens in the covered ChatGPT Enterprise and Business accounts in the EU by Sept. 30, 2026. It does not mean 55% of users, tasks or business value.

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

No. The figures cover OpenAI business accounts in the 27 EU member states, not all European companies or AI providers. They do not establish the share of businesses using agents across Europe.

Why might agents account for so many tokens?

Agent workflows can involve reading files, planning, generating material and checking results, which can produce more text than a short chat reply. OpenAI says agentic tasks can require substantially more computation.

What does the 8.1-fold small-business increase mean?

OpenAI reports that weekly active users of Codex or ChatGPT Work at businesses with 200 or fewer users rose 8.1 times from the first week of May to the week ending Oct. 5. The source material gives no absolute user count, and the baseline came before ChatGPT Work launched.

What remains unknown about the country comparisons?

The source material does not give account or user sample sizes for each country. Differences may reflect the mix of OpenAI customers in each market, so the percentages should not be treated as direct comparisons of national AI adoption or readiness.

Source: ThorstenMeyerAI.com