OpenAI’s Agentic AI Figure For EU Business Depends On The Denominator

🔍 Read the full analysis: OpenAI’s Agentic AI Figure For EU Business Depends On The Denominator 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 produced 55% of output tokens in its ChatGPT Enterprise and Business accounts across the 27 EU member states by Sept. 30, a figure that describes model output rather than how many tasks or users rely on agents. The company’s Oct. 14 report also recorded rapid growth in small-business use, but the data comes from OpenAI’s own customer base and does not establish how representative those accounts are of European businesses overall.

In Europe’s Agentic Opportunity: Moving from breadth to depth with AI agents, OpenAI Economic Research defines “agentic” use as activity through 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%.

OpenAI reported that weekly active users of either agent among businesses with 200 or fewer users grew 8.1 times between the first week of May and the week ending Oct. 5. Each EU member state recorded at least fivefold growth, the report said. Belgium led at 12.1 times, followed by Italy at 10.8 and Luxembourg at 10.0. OpenAI did not give absolute user counts for this growth measure.

Agentic output shares varied by country, from 85% in Lithuania and 81% in Cyprus to 22% in Slovenia and 26% in Austria. Germany recorded 51%, placing it 17th among the 27 member states. The report also said 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 the output per user of the EU frontier.

What the 55% Figure Measures

The headline share indicates where OpenAI’s model-generated text is being produced inside its EU business accounts. It does not show that agents handle most AI interactions, that most employees use them, or that the tools deliver most business value. OpenAI’s own Enterprise Signals page calls tokens an imperfect measure of business value and notes that agentic tasks can require substantially more computation.

That distinction matters because agent workflows may read files, plan, generate material, check results and repeat steps. A single run can produce much more output than a short chat response. The source author offers an illustrative calculation—not a finding from OpenAI’s report—that if agent runs generated 10 to 30 times the output of chat replies, a 55% token share could correspond to roughly 4% to 11% of interactions. The actual ratio is not supplied by the report.

For business leaders and policymakers, the figures point to growing use of agent products among OpenAI customers, but they cannot by themselves establish productivity gains or broad adoption across Europe. OpenAI’s policy argument is that the region should turn adoption into deeper use through skills, confidence and access to advanced tools. The published metrics do not measure whether those steps have raised output or improved business outcomes.

How OpenAI Built the Comparison

OpenAI’s Global Affairs team published the report on Oct. 14, 2026, alongside a global Enterprise Signals page. The analysis covers ChatGPT Enterprise and Business accounts in the EU’s 27 member states; it is not a survey of all European companies. Its measures track activity within the company’s products, so results reflect the customers and usage OpenAI can observe.

The timing affects both headline measures. Codex accounted for roughly one-fifth of EU output tokens in May and 34% by Sept. 30, according to the report’s chart. ChatGPT Work launched in July and reached 21% by the end of September. The report says a major part of the small-business user growth followed that launch. Thus the period’s increase includes the introduction of a new product, rather than measuring only gradual expansion of tools available throughout the full comparison window.

OpenAI’s report frames Europe’s opportunity as moving from broader adoption to deeper use. It also says EU companies make up about 10% of frontier firms worldwide, compared with about 65% in North America. Those figures use the report’s output-per-active-user definition of frontier firms; they are not a direct comparison of productivity, company quality or economic impact.

Limits in the EU Account Data

The report does not provide country-level account or user sample sizes, making it difficult to judge how much a small number of large customers may influence national shares. OpenAI itself notes that smaller countries including Cyprus, Estonia and Lithuania have relatively high agentic use within their larger firms. The country figures therefore describe observed OpenAI accounts, not national readiness rankings.

The report also gives no absolute number of small-business users behind the 8.1-times growth, and it does not separate the effects of the ChatGPT Work launch from other changes in product use. Nor does the 55% token share reveal the number of agent interactions, their outcomes, costs or value to customers. The source material also does not provide a complete account of how the global and EU frontier comparisons handle differences in customer mix.

As a result, it remains unclear how far these findings generalize beyond OpenAI customers, whether the growth will persist, and whether greater agent use is producing measurable business benefits. The report establishes a pattern in product activity, not an economy-wide result.

What Further Evidence Could Show

OpenAI’s stated policy direction is to help companies move from experiments toward deeper agent use, with attention to worker skills, confidence and access to advanced capabilities. The report does not set out a new timetable or a specific target for those measures. Its next evidentiary step would be clearer reporting on user counts, customer coverage and outcomes, alongside continued product-usage data.

Readers should also look for later reporting that makes the denominator explicit: whether a statistic counts tokens, interactions, active users, firms or completed tasks. Comparisons over time will be easier to interpret if they account for product launches and disclose country-level sample sizes. Until then, the current figures are best read as a snapshot of OpenAI’s EU business-account activity, rather than a measure of agent adoption across European business as a whole.

Key Questions

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

It means Codex and ChatGPT Work generated 55% of output tokens in the covered EU ChatGPT Enterprise and Business accounts by Sept. 30, 2026. It does not mean agents handled 55% of tasks or users’ AI interactions.

How did OpenAI define agentic use?

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

What does the 8.1-times growth figure cover?

It is the increase in weekly active users of Codex or ChatGPT Work among businesses with 200 or fewer users, from the first week of May to the week ending Oct. 5. OpenAI did not report the absolute user count, and it said a major part of the growth followed ChatGPT Work’s launch.

Does the report show that agents improved business productivity?

No. It reports product usage and output-token comparisons, not measured productivity gains or business value. OpenAI says tokens are an imperfect measure of business value.

Can the country figures rank EU countries by AI readiness?

Not reliably from the information provided. OpenAI did not publish country-level account or user sample sizes, and a small number of large customers could affect a country’s result. The figures describe OpenAI’s observed accounts, not all businesses in each country.

Source: ThorstenMeyerAI.com