How Much Agentic AI Is Used By EU Businesses? Check OpenAI’s Denominator

🔍 Read the full analysis: How Much Agentic AI Is Used By EU Businesses? Check OpenAI’s Denominator on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get the little things that make your day 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 its AI agents produced 55% of the output tokens generated in its ChatGPT Enterprise and Business accounts across the 27 EU member states by Sept. 30, a measure of activity inside the company’s customer base rather than a count of European businesses or tasks. The figures, published in a report on Oct. 14, also show fast growth in small-business use, while leaving unclear how widespread agent use is across the EU economy.

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

The company reports that weekly active users of Codex or ChatGPT Work at businesses with 200 or fewer users rose 8.1 times between May 1–7 and the week ending Oct. 5. OpenAI says every EU member state recorded at least fivefold growth over that period. Belgium had the highest reported growth at 12.1 times, followed by Italy at 10.8 and Luxembourg at 10.0. The report does not give absolute user counts.

Agentic output shares varied across countries, from 85% in Lithuania and 81% in Cyprus to 22% in Slovenia and 26% in Austria. Germany’s share was 51%, which the source places 17th among the 27 member states. OpenAI also says EU frontier firms—defined as the top 10% by output tokens per active user—generated about four times the tokens per user of typical EU firms in September. Non-EU frontier firms produced more than twice the output per user of EU frontier firms.

What the Token Share Reveals

The report offers a snapshot of how AI activity is changing within OpenAI’s EU business accounts, and it suggests that agent products are accounting for a substantial share of the computation those customers generate. That may matter to businesses planning software adoption, training and access to more advanced tools. It also gives policymakers a vendor-specific view of the divide between broad adoption and heavier use by leading firms.

But the headline figure does not mean that agents perform most AI tasks in Europe or that most companies have adopted them. Tokens measure generated text and other model output, not users, completed tasks or economic value. Agent runs can involve reading files, planning, producing work and checking results, generating many more tokens than a short chat response. OpenAI itself describes tokens as an imperfect measure of business value and says agentic tasks can require substantially more computation.

The distinction matters because a metric that rises with computation can make a product appear more prevalent than it is by number of interactions. The reported 55% is evidence about where output tokens go in this particular customer group. It does not establish how much time agents save, whether the work is better, or how many EU firms outside OpenAI’s customer base use similar systems.

A July Launch Shapes the Trend

The comparison starts in early May, before ChatGPT Work launched in July. According to the source article’s reading of OpenAI’s chart, Codex made up roughly a fifth of EU output tokens in May and 34% by Sept. 30. ChatGPT Work then contributed 21 percentage points by the end of September. The launch therefore accounts for a substantial part of the reported rise in agentic output, though the figures do not isolate the effect of the launch from other changes in customer use.

OpenAI’s small-business growth measure also begins before ChatGPT Work was available. The report says a major part of the increase followed the launch, but it does not state how many active users were counted at the start or end. An 8.1-fold increase from an undisclosed starting number cannot by itself show the scale of adoption across small firms in Europe.

OpenAI’s report is based on its own enterprise and business accounts, while its Enterprise Signals page supplies a global comparison. The company says EU firms account for about 10% of frontier firms worldwide, compared with about 65% in North America. These are company-defined categories and account data, not a census of all firms or AI systems operating in either region.

What the Data Cannot Establish

OpenAI does not disclose account or user sample sizes by country, making it difficult to judge how representative the country rankings are. The company notes that smaller markets, including Cyprus, Estonia and Lithuania, have relatively high agentic use within larger firms. A small number of large customers could therefore influence a country’s share; the published figures do not show how much.

The report also gives no absolute counts for the small-business users behind the 8.1-fold increase. It does not measure the share of interactions or jobs handled by agents, productivity gains, cost savings, accuracy, or the commercial value of agent output. Nor does it show whether the observed growth persisted after the reporting period. Those limits mean the findings cannot establish an EU-wide rate of business adoption or prove that Europe’s firms are closing the gap with global leaders.

The source article’s illustrative estimate that a 55% token share could correspond to a much smaller share of interactions assumes agent runs produce 10 to 30 times as much output as chat replies. That is an example based on an assumed ratio, not a figure from OpenAI’s report, and it should not be treated as an estimate of actual EU interaction shares.

The Next Test Is Broader Adoption

The next useful evidence will be whether agent use continues to grow after the initial period measured, and whether OpenAI or independent researchers publish absolute user counts, country-level sample sizes and task-based measures. Those details would help distinguish growth from a low starting base and show whether the country comparisons reflect broad participation or a concentrated set of customers.

OpenAI’s report calls for businesses to build worker skills and confidence and make advanced capabilities more accessible and affordable. The data released so far do not show whether those changes have occurred or what outcomes they have produced. For now, the report documents rising agent activity within OpenAI’s EU business accounts; its wider economic effects remain unmeasured.

Key Questions

What does OpenAI mean by 55% agentic use?

It means Codex and ChatGPT Work produced 55% of output tokens in the covered EU ChatGPT Enterprise and Business accounts by Sept. 30, 2026. It is not a claim that agents handled 55% of users, tasks or business activity.

How did OpenAI define an AI agent for this report?

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

Did small-business agent use rise 8.1 times across the EU economy?

No. OpenAI reports an 8.1-fold rise in weekly active users of Codex or ChatGPT Work at businesses with 200 or fewer users, within its covered accounts. It does not provide absolute counts or a measure of all small businesses in the EU.

Does the report show that EU businesses lag global firms?

OpenAI says its EU frontier firms produced about four times as many tokens per user as typical EU firms, while non-EU frontier firms produced more than twice as many as EU frontier firms. The comparison uses OpenAI’s account data and definition of frontier firms; it is not a census of companies worldwide.

Source: ThorstenMeyerAI.com