OpenAI’s EU Agentic AI Claim: Why The Counting Method Matters

🔍 Read the full analysis: OpenAI’s EU Agentic AI Claim: Why The Counting Method Matters on ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on tech for your team

• 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 October 14 that its AI agents generated 55% of output tokens in its ChatGPT Enterprise and Business accounts across the EU by September 30, a finding the company uses to argue that workplace AI use is becoming more agentic. The figure measures text generated by OpenAI’s tools—not the share of tasks, users or business value involving agents—so it describes activity within the company’s customer base, not the EU economy as a whole.

In its report Europe’s Agentic Opportunity: Moving from breadth to depth with AI agents, OpenAI Economic Research counts Codex, its coding agent, and ChatGPT Work, an agent mode that launched in July 2026, as agentic tools. By September 30, Codex accounted for 34% of output tokens in the covered EU accounts and ChatGPT Work for 21%. The remaining 45% came from standard ChatGPT use. The report covers accounts in the EU’s 27 member states.

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. Every member state recorded at least fivefold growth by this measure; Belgium, Italy and Luxembourg had the largest reported multiples. OpenAI did not provide absolute user counts alongside the growth figures.

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 states. OpenAI also compared high-use customers: its EU “frontier firms”—the top 10% by output tokens per active user—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 group.

What the 55% Share Actually Shows

The headline offers a measure of where generated text is going inside OpenAI’s business accounts. It does not establish that most AI interactions in those accounts, let alone across European workplaces, now involve agents. Agent workflows can read files, plan, generate code or documents, and check their work. Those steps can produce many more tokens than a short chat response, making token share a poor stand-in for the share of tasks.

OpenAI’s own Enterprise Signals page describes tokens as “an imperfect measure of business value” and says agentic tasks can require substantially more computation. The distinction matters for companies weighing adoption and for policymakers interpreting the report: high token use may indicate intensive agent activity, but it does not show whether the tools save time, improve output or deliver a return on investment.

The figures also come from a vendor’s own customers and include products that are new or recently introduced. They offer evidence of uptake within that customer group, but not an independent estimate of how widely European firms use AI agents. The report’s policy argument—that Europe should turn broad AI adoption into deeper use—should be read as OpenAI’s conclusion, not as a result established by the token metric alone.

New Tools Shape the Reported Growth

OpenAI released the report alongside a Global Affairs post titled The EU’s next chapter in AI is agentic. The company’s global Enterprise Signals page supplies additional comparisons. Together, the materials provide a detailed view of activity in ChatGPT Enterprise and Business accounts, but the measures remain specific to OpenAI products and customers.

Product timing is relevant to both headline measures. Codex’s share of EU output tokens rose from roughly a fifth in May to 34% by September 30, according to the report’s chart. ChatGPT Work launched in July and reached a 21% share by the end of September. OpenAI says a major part of the small-business user growth followed that launch. The 8.1-times comparison begins in the first week of May, before ChatGPT Work was available, so it combines adoption over time with the arrival of a new product.

OpenAI’s report defines frontier firms by output tokens per active user, not by revenue, productivity or an outside assessment of technical capability. It says EU firms account for about 10% of frontier firms worldwide, compared with about 65% in North America. These figures describe the company’s defined customer groups and should not be treated as a full census of global AI use.

What the Account Data Cannot Establish

The report does not disclose absolute account or user counts by country, making it difficult to judge how representative the national percentages are. OpenAI notes that smaller countries can have relatively high agentic use within larger firms. Without sample sizes, a small number of large, agent-heavy customers could have an outsized effect on a country’s result.

The figures also do not reveal how many tasks were handled by agents, how much time or money businesses saved, or whether the tools improved work quality. The report gives no baseline user totals for the 8.1-times small-business increase. A large growth multiple from an undisclosed starting point cannot by itself establish how widespread adoption is.

It remains unclear how closely OpenAI’s customer mix reflects businesses across the EU, including firms that do not use OpenAI products. The report’s data cannot settle whether Europe’s companies as a whole trail other regions in agent adoption or productivity. Nor does the supplied material provide an independent audit of the company’s account-level measures.

Evidence Needed Beyond Token Shares

The immediate next step is for readers and policymakers to distinguish OpenAI’s customer-level usage measures from economy-wide adoption claims. Further reporting could clarify the account counts behind country comparisons, the absolute number of small-business users and how much of the reported increase reflects ChatGPT Work’s launch.

To assess business impact, future data would need measures tied to work outcomes, such as tasks completed, time saved, quality or cost, alongside clear definitions and comparison groups. OpenAI’s report calls for stronger worker skills and confidence, as well as more accessible and affordable advanced tools. Whether that approach improves results for EU firms is not established by the current token and active-user figures.

Key Questions

What does OpenAI’s 55% figure measure?

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

Which tools does the report count as agents?

OpenAI counts Codex, its coding agent, and ChatGPT Work, an agent mode launched in July 2026. Standard ChatGPT accounts for the report’s remaining 45% of output tokens.

Does the 8.1-times increase mean small businesses widely adopted agents?

It means weekly active users of Codex or ChatGPT Work at businesses with 200 or fewer users were 8.1 times higher in the week ending October 5 than in the first week of May. OpenAI did not provide absolute user numbers, and says a major part of the growth followed ChatGPT Work’s launch.

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

Not on their own. The report gives country-level shares but no account or user sample sizes. The figures describe OpenAI’s customers in each country and may be affected by customer mix, especially in smaller markets.

Source: ThorstenMeyerAI.com