EU Business AI Is Mostly Agentic, Says OpenAI—But What’s Being Counted?

🔍 Read the full analysis: EU Business AI Is Mostly Agentic, Says OpenAI—But What’s Being Counted? on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get hardware and tech essentials 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 EU member states by Sept. 30, 2026. The figure, published in a new report on Oct. 14, describes the share of text generated by the company’s agent products—not the share of business tasks handled by agents or the value they create.

OpenAI’s Economic Research report, Europe’s Agentic Opportunity: Moving from breadth to depth with AI agents, counts two products as “agentic”: 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 and ChatGPT Work for 21%; standard ChatGPT accounted for 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 Oct. 5. OpenAI says every EU member state recorded 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 for this measure.

Agentic output shares varied across countries in OpenAI’s customer data. Lithuania had the highest reported share, at 85%, followed by Cyprus at 81%; Slovenia had the lowest, at 22%, and Austria was at 26%. Germany was at 51%, which the report ranks 17th among the 27 member states.

What the Token Share Shows

The report offers a view into how customers of one major AI provider are using its products, and it points to growing use of tools that can take multi-step actions rather than simply return a chat response. For businesses weighing how AI is being deployed, that is relevant evidence of increased agent activity within OpenAI’s own customer base.

But a token share is not a direct measure of workplace adoption or economic impact. An agent may read documents, plan, generate material, check its work and try again, producing far more output tokens than a short chat exchange. OpenAI’s global Enterprise Signals page describes tokens as an imperfect measure of business value and notes that agentic tasks can require substantially more computation. The 55% result therefore says where OpenAI’s generated output is concentrated, not that agents perform most work or deliver most value in European companies.

That distinction matters when comparing countries or assessing whether businesses have moved beyond experiments. OpenAI’s data covers its paying business accounts, not all EU firms or the full range of AI tools used across the region. The findings can inform discussion, but they cannot by themselves establish a Europe-wide rate of agent adoption.

How OpenAI Defines Agentic Use

The report was published by OpenAI’s Global Affairs team with a ten-page analysis from OpenAI Economic Research. Its underlying sample is ChatGPT Enterprise and ChatGPT Business accounts in EU member states. “Agentic” is defined through activity in Codex and ChatGPT Work, rather than as a general measure covering every AI agent or provider.

Product timing shapes the reported growth. Codex accounted for roughly a fifth of EU output tokens in May, according to the report’s chart, and reached 34% by Sept. 30. ChatGPT Work launched in July and represented 21% by the end of September. OpenAI says a major part of the small-business user growth followed that launch. The May-to-October comparison therefore spans the introduction of a product included in the measure, not only gradual changes in use of existing tools.

OpenAI also compares its highest-use customers with other firms. It defines “frontier firms” as the top 10% by output tokens per active user. In September, EU frontier firms produced about four times as many tokens per user as typical EU firms; frontier firms outside the EU produced more than twice the output per user of EU frontier firms. OpenAI says EU companies make up about 10% of frontier firms worldwide, compared with about 65% in North America. These comparisons use token output, not independently measured productivity or business results.

What the Dataset Cannot Establish

The report does not disclose the number of accounts or users behind each country’s figures, making it difficult to judge how representative country rankings are. OpenAI notes that smaller countries can have high agentic use within larger firms, but without sample sizes readers cannot tell how much individual customers influence the totals.

The reported 8.1-times rise is a relative change, and the source provides no absolute count of small-business users. It also does not show what share of interactions, tasks or business activity involved an agent. Because output-heavy agent runs can produce many more tokens than ordinary chat replies, the 55% token share cannot be converted into an interaction share without additional data. Any estimate based on an assumed difference in token volume would be a separate calculation, not a finding in OpenAI’s report.

Nor does the report establish that higher token use leads to better outcomes. It offers no direct comparison of productivity, cost savings, work quality or business performance. The available figures describe activity on OpenAI’s products and should not be read as a complete account of AI use across European businesses.

Evidence Needed Beyond Token Counts

OpenAI’s stated policy focus is to help businesses move from broader adoption to deeper use by building workers’ skills and confidence and making advanced capabilities more accessible and affordable. The report does not set out a timetable or specific targets for those steps.

Further reporting from OpenAI could clarify the picture by providing account and user counts, methodology details for country comparisons, and measures of tasks or outcomes alongside tokens. Until then, the next public milestone is additional data on how use develops after ChatGPT Work’s launch and whether the growth extends beyond OpenAI’s own customer base. The current report does not say when such data will be released.

Key Questions

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

It means Codex and ChatGPT Work generated 55% of output tokens in the covered ChatGPT Enterprise and Business accounts by Sept. 30, 2026. It does not mean agents handled 55% of tasks or created 55% of business value.

Which products does OpenAI count as agents?

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

How quickly did small-business agent use grow?

OpenAI reports 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 Oct. 5. It does not publish the underlying absolute user counts.

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

No. It covers OpenAI’s ChatGPT Enterprise and Business accounts, not all European businesses or all AI providers. Its 55% figure is based on output tokens, not the share of companies using agents.

Can the country rankings be treated as a measure of national AI readiness?

Not on the evidence provided. OpenAI does not report account or user sample sizes for each country, and its rankings reflect activity in its own customer base. The figures do not establish the overall level of AI readiness in each member state.

Source: ThorstenMeyerAI.com