🔍 Read the full analysis: Before You Cite OpenAI’s EU Agentic AI Claim, Read The Denominator on ThorstenMeyerAI.com
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
TL;DR
OpenAI’s October 14 report says Codex and ChatGPT Work generated 55% of output tokens in its EU business accounts by September 30. The figure is a measure of generated text, not the share of users, tasks or business value, and the report’s growth measures include a newly launched product. OpenAI does not publish country-level sample sizes or absolute small-business user counts.
OpenAI reported on October 14 that its AI agents generated 55% of output tokens in ChatGPT Enterprise and Business accounts across the European Union by September 30. The figure points to growing use of agent products among OpenAI’s business customers, but it does not mean agents handled most AI tasks or delivered most business value: it measures the volume of text generated by the tools.
OpenAI Economic Research’s ten-page report, Europe’s Agentic Opportunity: Moving from breadth to depth with AI agents, covers business accounts in the 27 EU member states. It counts Codex, OpenAI’s coding agent, and ChatGPT Work, an agent mode launched in July 2026, as agentic tools. By September 30, Codex accounted for 34% of output tokens and ChatGPT Work for 21%; standard ChatGPT accounted for the remaining 45%.
For businesses with 200 or fewer users, OpenAI says weekly active users of Codex or ChatGPT Work grew 8.1 times from the first week of May to the week ending October 5. Every member state recorded at least fivefold growth. Belgium led at 12.1 times, followed by Italy at 10.8 and Luxembourg at 10.0. OpenAI does not provide the absolute user counts behind that growth measure.
The country figures also vary widely: agentic tools’ share of output tokens ranged from 85% in Lithuania and 81% in Cyprus to 22% in Slovenia and 26% in Austria. Germany registered 51%, placing it 17th among the 27 states. OpenAI defines “frontier firms” as the top 10% of firms by output tokens per active user. In September, EU frontier firms produced about four times as many tokens per user as typical EU firms, while frontier firms outside the EU produced more than twice as many as the EU frontier.
OpenAI’s Europe report · in one view
From 22% to 85%: how agentic the EU’s OpenAI business accounts are
Share of output tokens produced by Codex and ChatGPT Work in September 2026, per member state, next to how fast small businesses adopted the agents since May.
Agentic share of output tokens, September 2026
Dark tick = EU median (56%). Right column: small-business agent users vs. May; green = at or above the median (8.1×).
- 1Lithuania85%6.8×
- 2Cyprus81%7.5×
- 3Hungary79%8.1×
- 4Poland76%6.5×
- 5Sweden73%8.0×
- 6Finland72%5.1×
- 7Estonia71%7.8×
- 8Netherlands68%8.5×
- 9Croatia68%6.8×
- 10Italy65%10.8×
- 11Slovakia58%7.8×
- 12Belgium57%12.1×
- 13Denmark56%9.9×
- 14Malta56%9.5×
- 15France55%9.2×
- 16Czechia52%8.1×
- 17Germany51%7.5×
- 18Romania49%5.9×
- 19Luxembourg47%10.0×
- 20Bulgaria42%5.9×
- 21Ireland39%9.5×
- 22Greece37%9.8×
- 23Latvia35%5.8×
- 24Spain30%8.8×
- 25Portugal29%6.5×
- 26Austria26%8.2×
- 27Slovenia22%9.2×
Read as a customer map, not a league table: no account or user counts per country. Share and growth barely relate (rank correlation -0.19): Slovenia is last on share but grows 9.2×, Lithuania first on share but grows 6.8×.
Source: OpenAI Economic Research, “Europe’s Agentic Opportunity: Moving from breadth to depth with AI agents” (October 2026); ChatGPT Enterprise and Business accounts. “Agentic” = Codex + ChatGPT Work (launched July 2026). Small and medium businesses = 200 or fewer associated users. OpenAI publishes no sample sizes per country. Rank correlation: own calculation (Spearman) from Figures 2 and 4.
Powered by Thorsten Meyer AI
The report offers a snapshot of activity inside OpenAI’s own business customer base, not a census of AI use across European companies. That distinction matters if the figures are used to support claims about the adoption of agents in the EU economy. OpenAI’s customers may not represent all businesses, and output-token share cannot by itself establish how many people use agents, how many tasks they complete or whether the work is valuable.
Token totals also reflect how products operate. An agent may read files, plan a task, generate an answer, check its work and repeat steps, producing far more text than a short chat exchange. OpenAI’s Enterprise Signals page describes tokens as an imperfect measure of business value and says agentic tasks can require substantially more computation. A 55% token share therefore indicates where model output is being generated; it is not evidence that agents account for 55% of interactions or economic impact.
The distinction is useful for policymakers and business leaders weighing investment in AI skills and infrastructure. OpenAI’s report argues that Europe should move from broad adoption to deeper use, but the available metrics cannot independently show whether companies are achieving productivity gains. They can describe product activity and suggest areas for further study, while leaving outcomes such as time saved, quality and returns unmeasured.
Products Worth Considering
Logitech MX Mechanical Wireless Illuminated Keyboard, Tactile Quiet, Graphite
Logitech MX Keys S Combo Wireless Keyboard and Mouse, Graphite
Apple iPad Pro 12.9-inch (2017), 512GB Wi-Fi, Space Gray
As an affiliate, we earn on qualifying purchases.
A July Launch Shapes the Growth
The timeline affects how to read both headline measures. OpenAI’s chart shows Codex accounted for roughly one-fifth of EU output tokens in May and 34% by September 30. ChatGPT Work launched in July and accounted for 21% by the end of September. That means the new product contributed more than half of the roughly 35-percentage-point rise in the agentic share described in the source analysis. The shift reflects adoption of a recently launched OpenAI tool as well as broader changes in use.
The 8.1-fold small-business figure compares weekly active users in the first week of May with the week ending October 5, a period that begins before ChatGPT Work existed. OpenAI says a major part of the growth followed its introduction. Growth from an undisclosed starting count can show rapid expansion without revealing how many businesses or people use the product in absolute terms.
Country rankings also need caution. OpenAI gives no account or user sample sizes by country, and says smaller states including Cyprus, Estonia and Lithuania have relatively high agentic use within their larger firms. The source analysis finds only a weak, slightly negative relationship between countries’ agentic-output shares and small-business growth rankings. The country figures may reflect differences in OpenAI’s customer mix, not a reliable ranking of national AI readiness.
The Missing Counts Behind the Claims
OpenAI has not published the absolute numbers of small-business users behind the 8.1-fold increase, or account and user counts for each country. Without those denominators, readers cannot determine how many people the growth represents or how representative the national percentages are. The report also does not provide an independent measure of task completion, productivity, cost savings or business value tied to agent use.
The 55% statistic is a token share, and the comparison with standard ChatGPT does not reveal the number of interactions or distinct tasks in each category. The source analysis offers an illustration: if an agent run generated 10 to 30 times the output of a chat reply, a 55% token share could correspond to roughly 4% to 11% of interactions. That is a hypothetical calculation based on an assumed ratio, not a figure reported by OpenAI.
It is also unclear how much the measures reflect changes in product availability, customer composition or sustained use over time. The figures cover OpenAI customers, not all EU businesses, and the report does not establish that its patterns generalize to the wider market.
What Later Data Could Show
The report’s next useful test will be whether later updates show continued agent use after the initial ChatGPT Work launch period, alongside absolute user counts and sample sizes. Comparisons across time would be easier to interpret if OpenAI disclosed how many accounts and active users sit behind each country and business-size category.
For claims about economic impact, readers would need measures beyond token output and account activity, such as tasks completed, time saved, costs and the quality of results. OpenAI’s report calls for wider access, workforce skills and deeper adoption, but it does not establish that these changes have already occurred across Europe. Until more outcome data is available, the strongest supported conclusion remains narrower: agents account for a large share of generated tokens in OpenAI’s EU business accounts, with rapid reported growth from a May baseline.
Key Questions
What does OpenAI’s 55% figure measure?
It measures the share of output tokens generated by Codex and ChatGPT Work in OpenAI’s EU Enterprise and Business accounts by September 30, 2026. It is not a share of users, tasks or business value.
How fast did small-business agent use grow?
OpenAI says weekly active users of Codex or ChatGPT Work at businesses with 200 or fewer users grew 8.1 times from the first week of May to the week ending October 5. The report does not give the absolute user counts.
Does the report show that most European AI tasks are now handled by agents?
No. The 55% statistic is based on generated token volume in OpenAI’s EU business accounts. The report does not give a share of tasks or interactions, and its customer data does not represent every European business.
Why does ChatGPT Work’s launch matter to the growth figures?
ChatGPT Work launched in July 2026, during the period covered by the report, and accounted for 21% of EU business-account output tokens by September 30. OpenAI says its launch contributed to small-business user growth, so the reported rise includes adoption of a new product.
Can the country rankings show which EU states are most ready for AI?
Not on their own. OpenAI does not publish country-level account or user sample sizes, and the reported figures may reflect the mix of its customers in each state. They should be read as OpenAI customer data, not a definitive ranking of national AI readiness.
Source: ThorstenMeyerAI.com
