The Measure Behind OpenAI’s Claim On Agentic AI In EU Business

🔍 Read the full analysis: The Measure Behind OpenAI’s Claim On Agentic AI In EU Business on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get the latest gadgets delivered free with Prime

• Fast, free delivery on millions of items
• Prime Video, Amazon Music and more included
• Member-only deals all year

Start your free Prime trial →Free trial for eligible customers · Cancel anytime

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 27 European Union member states by September 30, 2026. The figure, published in a report on October 14, describes activity among OpenAI’s own business customers; it does not show that agents perform most AI tasks or create most business value across Europe.

OpenAI Economic Research’s report, Europe’s Agentic Opportunity: Moving from Breadth to Depth with AI Agents, defines agentic use as activity from Codex, OpenAI’s coding agent, and ChatGPT Work, an agent mode launched in July 2026 that can work with files, browsers and company tools. The company says Codex represented 34% of EU account output tokens and ChatGPT Work 21% by September 30. Standard ChatGPT accounted for the remaining 45%.

For businesses with 200 or fewer users, OpenAI reports weekly active use of Codex or ChatGPT Work increased 8.1 times between the first week of May and the week ending October 5. Each member state recorded at least fivefold growth, according to the report. Belgium led at 12.1 times, followed by Italy at 10.8 and Luxembourg at 10.0. OpenAI does not provide absolute user counts alongside those growth figures.

Use varied sharply across countries in the company’s customer data: the agentic share ranged from 85% in Lithuania and 81% in Cyprus to 22% in Slovenia and 26% in Austria. Germany was at 51%, ranked 17th of 27 member states. These are shares of generated tokens, not national measures of adoption across all firms.

What the Token Share Captures

The figures offer a view of how quickly agent products are appearing in OpenAI’s business accounts, and the small-business growth measure suggests uptake rose after the products became available. But the token measure is not equivalent to the share of work handled by agents. An agent may read files, plan, generate material and revise it, producing far more output tokens than a short chat exchange. OpenAI itself describes tokens as an imperfect measure of business value and says agentic tasks can require substantially more computation.

That distinction matters for companies weighing adoption and for policymakers comparing European AI use with other regions. A larger token share may reflect longer or more computationally intensive runs, rather than more tasks, users or economic benefit. The report’s data can indicate where OpenAI’s computing activity is concentrated, but cannot establish whether companies are becoming more productive or whether workers are using agents for a majority of their work.

OpenAI also says EU frontier firms—defined as the top 10% by output tokens per active user—generated about four times as many tokens per user as typical EU firms in September. Frontier firms outside the EU generated more than twice as many as the EU frontier. The company estimates that EU businesses make up about 10% of frontier firms worldwide, compared with about 65% in North America. These measures describe intensity of use on OpenAI’s services, not independently verified rankings of productivity or AI capability.

How OpenAI Defines Agentic Use

The report was published alongside an OpenAI Global Affairs post titled The EU’s Next Chapter in AI Is Agentic. Its stated policy argument is that Europe should move from broad AI adoption toward deeper use of advanced tools, with more worker skills and confidence and greater access to capable, affordable systems. That is OpenAI’s interpretation of its customer data, rather than an independent assessment of EU-wide business readiness.

The timing affects how to read the growth numbers. Codex rose from roughly one-fifth of EU output tokens in May to 34% by September 30, according to the report’s chart. ChatGPT Work was introduced in July and accounted for 21% by the September cutoff. The report says a major part of small-business growth followed that launch. The May-to-October comparison therefore spans a new product release; it is not a like-for-like measure of demand for a stable set of tools.

The company’s report covers ChatGPT Enterprise and ChatGPT Business accounts in the EU, not every business or AI provider. Its country comparisons lack published account and user sample sizes, making it difficult to know how representative each national result is. OpenAI’s global Enterprise Signals material provides a broader company-wide framing, but does not turn these customer metrics into a census of European firms.

What the EU Figures Cannot Show

The report does not disclose absolute user counts for the small-business growth figure or account and user sample sizes for individual countries. Without those details, readers cannot tell how large the underlying populations were, how representative the country rankings are, or how much a small number of high-use customers influenced a national percentage.

It is also unclear how much of the increase reflects sustained changes in business practice rather than early use of newly launched products. The report supplies no independent measure of tasks completed, time saved, quality, revenue or productivity. Its findings are limited to OpenAI’s customer base and use of its services; they do not establish the prevalence of agents across the EU economy or across other AI platforms.

The source article offers an illustrative calculation that assumes an agent run produces 10 to 30 times the output of a chat reply, implying that a 55% token share could correspond to roughly 4% to 11% of interactions. That is the author’s hypothetical, not a figure from OpenAI’s report, and should not be treated as an estimate of actual interaction share.

Evidence Needed Beyond Token Counts

OpenAI’s immediate next step, according to its report, is to make advanced capabilities more accessible and affordable while building workers’ skills and confidence. The published figures run through September 30 for token shares and through October 5 for small-business weekly active use; the report does not announce a date for a follow-up release.

Further updates could show whether the growth continues after ChatGPT Work’s launch period, but a clearer account of business impact would require information beyond token volume. That could include longer measurement periods, user and account counts by country, and evidence about task completion or workplace outcomes. Until such data are available, the report is best read as a snapshot of agent use among OpenAI’s EU business customers, not a verdict on the state of agentic AI across European business.

Key Questions

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

OpenAI says Codex and ChatGPT Work produced 55% of output tokens in its EU ChatGPT Enterprise and Business accounts by September 30, 2026. Tokens measure generated text and other output, not the proportion of users, tasks or business value attributable to agents.

Which products count as agentic in the report?

The report counts Codex, OpenAI’s coding agent, and ChatGPT Work, an agent mode launched in July 2026 that can use files, browsers and company tools.

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

No. It covers OpenAI’s ChatGPT Enterprise and Business accounts in the EU. It does not measure all European companies or users of other AI services.

What does the 8.1-times small-business growth figure mean?

OpenAI reports that weekly active use of Codex or ChatGPT Work among businesses with 200 or fewer users was 8.1 times higher in the week ending October 5 than in the first week of May. The report does not give absolute user counts, and the period includes ChatGPT Work’s July launch.

Does the report measure productivity gains?

No. It reports product use and output tokens, but does not establish effects on productivity, quality, time saved or revenue. OpenAI says tokens are an imperfect measure of business value.

Source: ThorstenMeyerAI.com