🔍 Read the full analysis: The Numbers Behind OpenAI’s Claim About Agentic AI In EU Business on ThorstenMeyerAI.com
Get audio and creator gear 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 its Codex and ChatGPT Work agents generated 55% of output tokens from ChatGPT Enterprise and Business accounts in the EU by September 30. The figure measures generated text, not the share of tasks, users or business value; rapid growth also followed ChatGPT Work’s July launch. OpenAI’s customer data does not establish adoption across the broader European economy.
OpenAI says its Codex and ChatGPT Work agents generated 55% of output tokens in its ChatGPT Enterprise and Business accounts across the EU by September 30, according to a report published October 14. The figure offers a view of activity among OpenAI’s business customers, but it measures generated text—not the share of work tasks handled by agents or the business value they deliver.
OpenAI Economic Research’s report, Europe’s Agentic Opportunity: Moving from breadth to depth with AI agents, covers business accounts in the EU’s 27 member states. OpenAI defines agentic output as material generated by Codex, its coding agent, and ChatGPT Work, an agent mode introduced in July 2026 that can work with files, browsers and company tools. In the September 30 snapshot, Codex accounted for 34% of output tokens and ChatGPT Work for 21%. Standard ChatGPT made up the remaining 45%.
OpenAI also reports that 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 October 5. Each EU member state recorded at least fivefold growth, the report says. Belgium had the highest reported increase at 12.1 times, followed by Italy at 10.8 times and Luxembourg at 10 times. OpenAI does not provide absolute user counts alongside those growth multiples.
The report shows wide differences in the share of output attributed to agents: Lithuania was at 85% and Cyprus at 81%, compared with 22% in Slovenia and 26% in Austria. Germany’s share was 51%, placing it 17th among the 27 countries. OpenAI also compares high-output businesses: 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. Frontier firms outside the EU generated more than twice the output per user of the EU frontier group.
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 figures suggest that agent products are becoming a substantial part of activity in OpenAI’s EU business accounts. That matters to companies deciding whether to test or expand tools that can carry out multi-step work, and to policymakers weighing how adoption, workforce skills and access to advanced AI differ across Europe.
But the headline share is not a measure of how much work agents complete. OpenAI’s Enterprise Signals page describes tokens as an imperfect measure of business value and says agentic tasks can require substantially more computation. An agent may inspect files, plan, generate content and revise it, producing more tokens than a short chat reply. The report therefore supports a claim about where output-token volume is generated within OpenAI’s accounts—not that agents perform most tasks, serve most users or create most value.
The growth figures also need a product-launch context. ChatGPT Work was not available during the May baseline period. Its introduction in July contributed 21 percentage points of the 55% September agentic share, while Codex’s share rose from roughly one-fifth in May to 34% by September, according to the chart described in the report. The increase reflects both uptake and the addition of a new product.
Products Worth Considering
Logitech MX Keys Combo for Business Keyboard and Mouse – Graphite
Bose SoundLink Flex Bluetooth Speaker (2nd Gen), Black
Anker SOLIX S2000 Portable Power Station, 2,010Wh, 1,500W
As an affiliate, we earn on qualifying purchases.
A Snapshot of OpenAI Customers
The report is based on ChatGPT Enterprise and ChatGPT Business accounts, not a survey of every EU company. Its findings describe activity within OpenAI’s own customer base, which means they cannot by themselves establish the rate of agent use across Europe’s business sector. The company’s global Enterprise Signals page provides a wider comparison, while the new report focuses on EU usage and differences among member states.
OpenAI’s policy conclusion is that Europe should move from “broader adoption” to “deeper use” by supporting worker skills and confidence and making advanced capabilities more accessible and affordable. That recommendation follows from the report’s comparison between EU frontier firms and peers outside the bloc. The figures do not identify why the gap exists or show whether it results from access, investment, industry mix, company size or other factors.
The country results should also be read as customer-base measurements rather than a ranking of national readiness. OpenAI provides no country-level account or user sample sizes in the material supplied. It notes that smaller countries including Cyprus, Estonia and Lithuania have relatively high agent use within larger firms. Without sample counts and a fuller account of customer composition, the national percentages may be sensitive to which companies use OpenAI products.
What the EU Figures Leave Out
The report does not provide absolute user counts, country sample sizes or a measure of tasks completed. It also does not quantify productivity gains, cost savings or other business outcomes. Those omissions make it difficult to judge how representative the country comparisons are or how much practical value the reported activity produced.
The 8.1-times growth rate is measured against the first week of May, before ChatGPT Work launched, and OpenAI supplies no baseline count in the source material. It is therefore a growth multiple, not evidence that a large share of small businesses now use agents. Nor does the 55% token share establish that most AI interactions or company workflows are agentic.
OpenAI’s results cover its own products and customers. The data do not show whether companies using other AI providers have similar patterns, how the figures compare with all EU businesses, or what factors explain the gap between EU frontier firms and those outside the bloc. The report’s numbers are attributable to OpenAI and have not, in the supplied material, been independently verified.
Evidence Needed Beyond OpenAI Accounts
The immediate next step is for readers and decision-makers to distinguish product activity from economy-wide adoption. More useful comparisons would include absolute active-user counts, country-level sample sizes, the number and type of tasks completed, and results tied to costs or productivity. OpenAI’s report does not say when it will publish those measures.
Further data over time could show whether agent use continues to grow after ChatGPT Work’s launch effect is separated from the May baseline. Comparisons across providers and independent studies of business outcomes would also help establish whether the reported usage patterns reflect a wider shift in EU workplaces. For now, the report is a detailed snapshot of OpenAI’s business accounts, with important limits on what it can say about European companies overall.
Key Questions
What does OpenAI mean by 55% of AI output?
It means Codex and ChatGPT Work generated 55% of output tokens in OpenAI’s EU ChatGPT Enterprise and Business accounts by September 30, 2026. It does not mean agents handled 55% of users’ tasks or created 55% of business value.
Which products did OpenAI count as agents?
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.
How did small-business agent use change?
OpenAI says 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 October 5. The report gives no absolute user count, and the comparison begins before ChatGPT Work launched.
Do the country figures show which EU nations are best at AI?
No. They describe agent output shares among OpenAI’s customers in each country. OpenAI does not provide country-level sample sizes in the supplied material, so the figures are not a reliable ranking of national AI readiness.
Does the report show that agents improve business performance?
Not directly. The reported measures cover output tokens and weekly active users, not productivity, revenue, cost savings or task completion. OpenAI itself cautions that tokens are an imperfect measure of business value.
Source: ThorstenMeyerAI.com
