Does The Denominator Support OpenAI’s EU Business AI Headline?

🔍 Read the full analysis: Does The Denominator Support OpenAI’s EU Business AI Headline? 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 says its agent products generated 55% of output tokens in its ChatGPT Enterprise and Business accounts across the 27 EU member states by September 30, a figure that signals growing use of its newer tools but does not measure the share of tasks or business value they produce. The finding comes from a report published October 14 by OpenAI Economic Research and describes OpenAI’s own business customers, not all European companies.

OpenAI’s report, Europe’s Agentic Opportunity: Moving from breadth to depth with AI agents, counts output from two products as “agentic”: Codex, its coding agent, and ChatGPT Work, an agent mode that can work with files, browsers and company tools. By the end of September, Codex accounted for 34% of output tokens and ChatGPT Work for 21%; conventional ChatGPT accounted for the remaining 45%, according to the report.

The report also says 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. Each EU member state recorded at least fivefold growth on that measure. Belgium, Italy and Luxembourg had the highest reported growth, at 12.1 times, 10.8 times and 10.0 times, respectively. OpenAI does not provide absolute user counts alongside these growth figures.

Agentic output shares varied across countries in OpenAI’s customer data, from 85% in Lithuania and 81% in Cyprus to 22% in Slovenia and 26% in Austria. Germany was at 51%, ranking 17th among the 27 states. The report also defines “frontier firms” as the top 10% of firms by output tokens per active user: EU frontier firms generated about four times the output per user of typical EU firms, while frontier firms outside the EU generated more than twice the output of the EU frontier group.

What the 55% Measures

The headline is useful as a snapshot of where computation is going inside OpenAI’s paid business accounts. It is not a direct measure of how often workers choose agents, how many work tasks agents complete, or whether those tasks deliver better results. Output tokens are generated text, and an agent may produce much more of it than a short chat exchange while reading files, planning, drafting and revising.

OpenAI’s Enterprise Signals page describes tokens as an imperfect measure of business value and says agentic tasks can require substantially more computation. That caveat changes how readers should interpret the 55%: the figure supports the claim that agents account for more of the measured output in this customer set, but it cannot establish that most AI interactions or economic benefits are agent-driven.

The distinction matters to companies and policymakers using these figures to judge adoption. A high token share may reflect intensive use by some customers, the nature of agent workflows, or both. Without measures of completed tasks, outcomes, user counts and costs, the report cannot show how far European businesses have moved from experimentation to productive deployment.

A Fast Rise After Product Launch

The reported period combines a growing Codex share with the introduction of ChatGPT Work. According to the report’s chart, Codex accounted for roughly one-fifth of EU output tokens in May and 34% by September 30. ChatGPT Work launched in July and reached a 21% share by the end of September. OpenAI says a major part of the rise in small-business agent use followed that launch.

This timing is material to the 8.1-times growth claim. Its baseline is the first week of May, before ChatGPT Work existed, and the report supplies no absolute number of active users. The multiple therefore shows a sharp rise from that starting point among the covered accounts, but does not on its own establish how many businesses adopted the tools or whether usage will continue at the same pace.

OpenAI’s policy argument is that Europe should turn broader AI adoption into deeper use by developing worker skills and confidence and making advanced capabilities more accessible and affordable. The report says EU companies make up about 10% of global frontier firms, compared with about 65% in North America. These comparisons are drawn from OpenAI’s own accounts and its definition based on output tokens per active user.

Limits in OpenAI’s EU Data

The report does not publish country-level account or user counts, making it difficult to judge how representative the national comparisons are. OpenAI notes that smaller countries, including Cyprus, Estonia and Lithuania, have relatively high agentic use within larger firms. Without sample sizes, a small number of customers with intensive agent use could have a large effect on a country’s share.

Nor does the report provide an absolute count for the small-business active-user increase or a comparison with adoption across the wider EU economy. The May baseline predates ChatGPT Work’s launch, so product availability is one factor in the subsequent growth. The supplied material also does not establish whether token use translates into completed work, cost savings, improved quality or lasting adoption.

As a result, the figures should be read as usage measures for OpenAI accounts, not as a census of European business AI. The report does not establish why EU frontier firms trail their non-EU counterparts, or whether the difference reflects access, company mix, customer behavior or other factors.

Evidence Needed Beyond Token Shares

The October report offers a snapshot through September 30, with the small-business growth measure extending through the week ending October 5. No later update or independent validation is specified in the source material. Readers should look for future reporting that gives absolute user and account counts, explains country-level sample sizes, and tracks adoption over periods that are not anchored to a new product launch.

Further evidence on completed tasks, business outcomes, costs and the spread of use across company sizes would help show whether increased agent output corresponds to deeper adoption. Until those measures are available, the confirmed development is narrower: OpenAI reports that its agents generated most of the output tokens in its covered EU business accounts 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 ChatGPT Enterprise and Business accounts across the EU by September 30. It is not a count of tasks, users or business value.

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

No. The figures cover OpenAI’s business accounts in the 27 EU member states, not every European company or AI provider’s customers. The report does not give an EU-wide adoption rate.

How did small-business use change?

OpenAI reports an 8.1-times rise in 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. It does not provide the absolute user counts, and says a major part of the growth followed ChatGPT Work’s launch.

Why might token share overstate how common agent use is?

Agent workflows can generate more text than a short chat response because they may read materials, plan, draft and revise. A large share of tokens can therefore come from a smaller share of interactions. The report does not quantify the relationship between token volume and interaction counts.

What remains unknown about the country rankings?

OpenAI does not publish the number of accounts or users in each country. Without those sample sizes, readers cannot determine how representative the country-level shares are or how much a few large customers may affect them.

Source: ThorstenMeyerAI.com