OpenAI’s Agentic AI Claim In Europe: Read Beyond The Percentage

🔍 Read the full analysis: OpenAI’s Agentic AI Claim In Europe: Read Beyond The Percentage on ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on office and shipping supplies

• Business-only prices and quantity discounts
• Tax-exempt purchasing
• Multiple users, one account, clear invoices

Create a free Business account →Free to register

As an affiliate, we earn on qualifying purchases.Hide

OpenAI said Codex and ChatGPT Work produced 55% of output tokens in its ChatGPT Enterprise and Business accounts across the 27 EU member states by September 30, according to a report published October 14. The figure points to growing use of the company’s agents, but it measures generated text—not the share of users, work tasks or business value handled by agents.

In Europe’s Agentic Opportunity: Moving from breadth to depth with AI agents, OpenAI Economic Research defines “agentic” use as activity with Codex, its coding agent, and ChatGPT Work, an agent mode launched in July 2026 that can work with files, browsers and company tools. Codex accounted for 34% of EU account output tokens by September 30, while ChatGPT Work accounted for 21%. Plain ChatGPT made up the remaining 45%.

OpenAI also reported that weekly active users of Codex or ChatGPT Work at businesses with 200 or fewer users rose 8.1 times between the first week of May and the week ending October 5. The company said every EU member state grew at least fivefold; Belgium, Italy and Luxembourg recorded the highest reported multipliers. The report does not provide the underlying absolute user counts.

Geographic and company-size comparisons offer another view of the data. Agentic tools made up 85% of output tokens in Lithuania and 81% in Cyprus, compared with 22% in Slovenia and 26% in Austria; Germany’s share was 51%. OpenAI defines “frontier firms” as the top 10% of firms by output tokens per active user. In September, EU frontier firms produced about four times the output tokens per user of typical EU firms, while frontier firms outside the EU produced more than twice as many as the EU frontier.

What the Token Share Shows

The report offers a detailed look at activity inside OpenAI’s own paid business products, and its numbers suggest that agent tools are taking up a growing share of the computing those customers use. But more tokens do not establish more tasks or more economic value. An agent may read files, plan, generate material and check its work, producing much more output than a short chat exchange. OpenAI’s Enterprise Signals page itself calls tokens an imperfect measure of business value and says agentic tasks can require substantially more computation.

That distinction matters for business leaders and policymakers deciding whether adoption is changing workplace productivity. The reported 55% is a measure of output, not evidence that agents now perform most work in European companies. OpenAI’s policy argument is that Europe should move from broad adoption toward deeper use, with more worker skills and confidence and wider access to advanced tools. The report can inform that discussion, but its product-usage data cannot, on its own, show whether the tools improve productivity or how benefits are distributed.

The company’s figures also describe its customer base rather than the whole European economy. They exclude businesses that do not use ChatGPT Enterprise or Business, and do not establish how representative participating accounts are. The findings are useful as a window into one vendor’s products, not as a direct census of AI use across Europe.

How OpenAI Counts Agent Use

OpenAI published the report through its Economic Research team alongside a Global Affairs post on October 14. Its global Enterprise Signals page supplies wider company context, but the EU report’s principal figures concern ChatGPT Enterprise and Business accounts in the 27 EU countries. The measures cover different periods: the output-token share is dated September 30, while small-business weekly active-user growth runs through the week ending October 5.

The product timeline affects how the trend should be read. OpenAI’s chart shows Codex accounting for roughly one-fifth of EU output tokens in May and 34% by September 30. ChatGPT Work launched in July and reached 21% by the end of September. OpenAI says a major part of small-business user growth followed that launch. The increase therefore tracks both adoption and the addition of a new product to the measure; it does not isolate an underlying change in business practices.

OpenAI CFO Sarah Friar described the finding as a change in how its customers use AI, saying more than half of AI use among EU enterprise and business customers “now involves these tools.” The report’s more specific measure is output tokens. That distinction is important when interpreting the broader wording of the statement.

What the Data Cannot Establish

The report does not disclose account or user sample sizes by country, making it difficult to judge how much a small number of large customers may affect national percentages. OpenAI notes that some smaller countries have relatively high agentic use within larger firms. The country rankings should therefore be treated as figures for the company’s customer base, not definitive rankings of national readiness.

The report also gives no absolute user totals for the 8.1-fold small-business increase. The comparison begins in the first week of May, before ChatGPT Work launched, and the source says the launch contributed substantially to growth. Without the starting count, the multiplier does not show how many businesses or people adopted the tools, nor does it establish a broad shift across Europe’s small-business sector.

Finally, the supplied figures do not show what proportion of interactions were agentic, whether agent use improved work outcomes, or how much value customers gained. A hypothetical conversion from token share to interaction share would depend on assumptions about the relative output of agent runs and chat replies; it is not a result reported by OpenAI. These questions remain open.

What OpenAI’s Next Data Should Show

OpenAI’s October report sets out a case for Europe to turn experimentation into deeper use, but it does not announce a specific next reporting date or measurement milestone. Further data would be needed to test whether the trends continue beyond the period covered and whether growth persists after the initial launch effect of ChatGPT Work.

For readers assessing future updates, the most useful additions would include absolute user and account counts, country-level sample sizes, and consistent comparisons across products and time. Measures tied to completed tasks, outcomes or independently assessed productivity would also help distinguish increased computation from increased business value. Until then, the report’s clearest confirmed finding is limited to OpenAI’s own accounts: its two named agent products generated a majority of their output tokens by the end of September.

Key Questions

What does OpenAI mean by agents producing 55% of output?

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

How did OpenAI calculate the small-business growth figure?

OpenAI reported 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 October 5. The report does not give absolute user counts, and says a major part of growth followed ChatGPT Work’s launch.

Does the report cover all European businesses?

No. Its figures cover ChatGPT Enterprise and Business accounts in the 27 EU member states. They do not represent every European business or every AI provider.

Which EU countries had the highest agentic output share?

OpenAI reported the highest shares in Lithuania at 85% and Cyprus at 81%. It did not provide country-level account or user sample sizes, so the rankings may reflect differences in the customer base.

Source: ThorstenMeyerAI.com