OpenAI’s Claim On EU Agentic AI: The Headline And The Denominator

🔍 Read the full analysis: OpenAI’s Claim On EU Agentic AI: The Headline And The Denominator on ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on monitors, keyboards and dev gear

• 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 generated 55% of output tokens in its ChatGPT Enterprise and Business accounts across the European Union by September 30, 2026. The finding, published in a new report on October 14, describes how much text the company’s AI systems produced for its business customers; it does not show that agents handled most AI tasks or created most business value in Europe.

OpenAI Economic Research’s report, Europe’s Agentic Opportunity: Moving from Breadth to Depth with AI Agents, covers business accounts in the 27 EU member states. It classifies Codex, OpenAI’s coding agent, and ChatGPT Work, an agent mode launched in July, as “agentic” tools. By September 30, Codex represented 34% of output tokens and ChatGPT Work 21%; standard ChatGPT made up the remaining 45%.

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. OpenAI says every member state saw at least fivefold growth. Belgium, Italy and Luxembourg had the largest reported increases, at 12.1, 10.8 and 10.0 times, respectively. The report does not provide absolute user counts alongside those growth figures.

Agentic output shares varied widely across countries: Lithuania was reported at 85% and Cyprus at 81%, while Slovenia was at 22% and Austria at 26%. Germany’s reported share was 51%, placing it 17th among the 27 member states. OpenAI also compared its highest-output users with other customers: EU “frontier firms,” defined as the top 10% by output tokens per active user, produced about four times the output of typical EU firms. Firms outside the EU at the frontier produced more than twice the output of the EU frontier group.

What the Token Share Measures

The 55% figure is a measure of generated output, not a direct measure of how widely agents are used. Agent workflows can generate many tokens while reading files, planning, producing work and checking results. A smaller number of long agent runs could therefore account for a large share of output even if most interactions were still ordinary chat.

That distinction matters for policymakers and businesses deciding whether the figures show a broad change in work. OpenAI’s report argues that Europe should turn adoption into deeper use by improving worker skills and confidence and making advanced tools more accessible and affordable. The data may help describe activity among the company’s business customers, but it cannot by itself establish economy-wide adoption, productivity gains or return on investment.

The figures also come from a vendor’s own customer base. They offer a useful view of product use within those accounts, but not a representative survey of all European firms or of businesses using competing systems. The report’s output-based comparison between frontier firms likewise indicates token intensity, not which firms get the most economic value from AI.

How OpenAI Defines Agent Use

OpenAI published the report through its Economic Research team alongside a post from its Global Affairs team, titled The EU’s Next Chapter in AI Is Agentic. Its separate Enterprise Signals page provides a global view. Together, the material presents OpenAI’s account of agent use among its enterprise and business customers, rather than an independent census of European workplaces.

The comparison period is important. ChatGPT Work launched in July 2026, while the small-business growth calculation starts in the first week of May. OpenAI’s report says a major part of the increase followed that launch. The published figures therefore combine growing use of Codex with adoption of a newly introduced product; they do not isolate longer-term changes in business practice from the effects of product availability.

OpenAI’s worldwide analysis defines frontier firms by output tokens per active user. In the report’s September comparison, EU companies accounted for about 10% of frontier firms globally, compared with about 65% in North America. Those are figures about OpenAI’s customer activity under its chosen definition, not a ranking of all companies or a direct measure of national AI capability.

What the EU Figures Cannot Show

The report does not provide the number of accounts or users in each country, making it difficult to judge how representative the country-level percentages are. OpenAI notes that smaller countries such as Cyprus, Estonia and Lithuania have relatively high agentic use within larger firms. Without country-level sample sizes, the rankings may reflect the particular mix of customers in the data rather than differences across national economies.

The report also does not give absolute user counts for the small-business growth measure or separate the effect of ChatGPT Work’s launch from other causes of increased activity. The reported figures do not establish how many interactions were agentic, how much work was completed, or whether agent use improved productivity. Those outcomes remain unmeasured in the material described here.

Evidence Needed Beyond Output Tokens

OpenAI’s report points to worker skills, confidence and access to advanced tools as areas for further development, but it does not set out a new measurement timetable or a specific next milestone. A clearer picture of adoption would require absolute user and account counts, country-level sample sizes, and measures that distinguish tasks and outcomes from the volume of generated text.

For now, the figures should be read as a snapshot of OpenAI’s EU business accounts through September and early October 2026. Whether the growth persists after the initial ChatGPT Work launch period, and whether it leads to measurable business gains, remains to be established.

Key Questions

What does OpenAI mean by saying agents produced 55% of output?

It means Codex and ChatGPT Work together accounted for 55% of the output tokens generated in the covered EU ChatGPT Enterprise and Business accounts by September 30, 2026. It is not a claim that agents handled 55% of users’ tasks or delivered 55% of business value.

Which tools did OpenAI count as agentic?

The report counts Codex, OpenAI’s coding agent, and ChatGPT Work, an agent mode launched in July 2026. Standard ChatGPT was counted separately.

How did small-business agent use change?

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 provide absolute user numbers in the cited material, and says a major part of the growth followed ChatGPT Work’s launch.

Does the report show that European businesses are more productive?

No. The report measures product activity, including output tokens and weekly active users. It does not establish productivity gains, task completion rates or return on investment.

Source: ThorstenMeyerAI.com