OpenAI Says Most EU Business AI Is Agentic—The Denominator Tells More

🔍 Read the full analysis: OpenAI Says Most EU Business AI Is Agentic—The Denominator Tells More on ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on monitors, keyboards and dev gear

• Business-only prices and quantity discounts
• Tax-exempt purchasing
• Multiple users, one account, clear invoices

Create a free Business account →Free to register

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 ChatGPT Business accounts across the European Union by September 30, a finding the company uses to argue that workplace AI use is becoming more agentic. The figure covers OpenAI’s own business customers and measures generated text, not how many employees, tasks or business outcomes involve agents.

In a report published October 14 by OpenAI Economic Research, the company defined “agentic” use as activity involving Codex, its coding agent, and ChatGPT Work, an agent mode launched in July 2026 that can work with files, browsers and company tools. Codex accounted for 34% of output tokens and ChatGPT Work for 21%; ordinary ChatGPT accounted for the remaining 45%. The data covers accounts in the 27 EU member states.

OpenAI also reported fast growth in smaller business accounts, defined as businesses with 200 or fewer users. Weekly active users of Codex or ChatGPT Work grew 8.1 times between the first week of May and the week ending October 5. Every member state recorded at least fivefold growth, the report said. Belgium had the largest reported increase, at 12.1 times, followed by Italy at 10.8 times and Luxembourg at 10 times.

The company’s country figures vary widely: agents’ share of output ranged from 85% in Lithuania and 81% in Cyprus to 22% in Slovenia and 26% in Austria. Germany was at 51%, ranking 17th among the 27 states. OpenAI separately said EU frontier firms—its top 10% of firms by output tokens per active user—generated about four times the tokens per user of typical EU firms. Frontier firms outside the EU generated more than twice the EU frontier rate.

What the Token Share Shows

The report offers a snapshot of how some companies are using OpenAI’s newer tools, and points to increased agent activity within its business customer base. That may matter to employers planning software access, staff training and workplace workflows. But a token share is not a direct measure of adoption across Europe’s economy or of productivity gains.

Agents can perform multi-step work, such as reading files, planning, generating code or documents, and checking results. Such runs can generate substantially more text than a short chat response. OpenAI itself describes tokens as an imperfect measure of business value and says agentic tasks can require more computation. The 55% figure therefore indicates where output tokens were produced in these accounts; it does not establish that agents handled most business tasks or created most value.

The company’s report also frames the gap between EU frontier firms and those outside the bloc as a reason to expand deeper use. OpenAI says EU firms make up about 10% of frontier firms worldwide, compared with about 65% in North America. That comparison is based on the report’s token-per-user definition of frontier use, not an independent measure of competitiveness or overall AI capability.

How OpenAI Defined Agent Use

The October 14 publication consists of an OpenAI Global Affairs post, “The EU’s next chapter in AI is agentic,” and a ten-page OpenAI Economic Research report, “Europe’s Agentic Opportunity: Moving from breadth to depth with AI agents.” OpenAI’s Enterprise Signals page provides a global comparison. The figures concern ChatGPT Enterprise and Business accounts, not all European companies using AI products from any provider.

The timing affects how to read the growth figures. Codex already contributed roughly a fifth of EU account output tokens in May, according to the chart described in the report. ChatGPT Work launched in July and accounted for 21% by September 30. The source article’s reading of the chart is that Work contributed more than half of the roughly 35-point increase in agentic token share. The May baseline for small-business growth also predates that launch; OpenAI says a major part of the subsequent increase followed Work’s introduction.

OpenAI’s stated policy takeaway is that Europe should turn broad AI adoption into deeper use by building workers’ skills and confidence and making advanced tools more accessible and affordable. That is the company’s interpretation and recommendation, rather than a finding that the reported usage has already improved business performance.

What the Dataset Leaves Open

The report does not disclose absolute user counts or account counts for the small-business growth measure, nor country-level sample sizes for the national comparisons. Without those figures, readers cannot judge how many businesses the changes represent or how representative each country’s result is. OpenAI notes that smaller countries can show high agent use within larger firms, but the published chart does not let readers test how much a few large customers affect each national share.

The 8.1-times increase is relative to the first week of May, before ChatGPT Work launched. The source does not give the absolute baseline needed to assess the scale of the user increase. Nor does the report’s token measure identify the number of agentic interactions: a longer agent run can produce many more tokens than a chat exchange. Any estimate of what share of interactions was agentic would require assumptions not supplied by the report.

OpenAI is reporting on customers using its own services, and the data is not an independent survey of European businesses. The report does not establish whether agent use improved productivity, reduced costs or produced better work. Those outcomes, and whether the reported patterns extend to other AI providers or firms that do not use OpenAI, remain unanswered by these figures.

What to Watch in Future Data

The next useful evidence would include updated figures showing absolute user and account counts, country sample sizes, and comparable measures over time. Those details could help distinguish broad adoption from growth concentrated among a smaller set of customers, and show whether use continued after ChatGPT Work’s launch period.

Readers should also watch for data that measures tasks and outcomes alongside token volumes, such as the kinds of work agents handle and whether businesses report changes in time, cost or quality. OpenAI’s report calls for more skills and access to advanced tools, but it does not set out a specific next measurement date or independently verified outcome study. For now, the confirmed finding is narrower: agents generated a majority of output tokens in the company’s covered EU business accounts by September 30.

Key Questions

What does OpenAI mean by “55% of output”?

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

How did small-business agent use change?

OpenAI reported that weekly active users of Codex or ChatGPT Work among 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 provide the absolute user count.

Did agent use grow evenly across EU countries?

No. OpenAI reported agentic output shares from 22% in Slovenia to 85% in Lithuania. It did not publish country-level account or user counts, so the results should not be treated as a definitive ranking of national AI adoption.

Does the report show that agents improved productivity?

No. It reports usage and token-volume measures, not independently verified productivity, cost or quality outcomes. OpenAI also cautions that tokens are an imperfect measure of business value.

Source: ThorstenMeyerAI.com