OpenAI’s Claim On Agentic AI In EU Business, Explained Through The Numbers

🔍 Read the full analysis: OpenAI’s Claim On Agentic AI In EU Business, Explained Through The Numbers 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 reported that its coding and workplace agents generated 55% of output tokens in its ChatGPT Enterprise and Business accounts across the 27 EU member states by Sept. 30, 2026. The figure comes from a company report published Oct. 14; it measures generated text in OpenAI’s own business products, not the share of European companies, tasks or business value attributable to agents.

OpenAI Economic Research’s report, Europe’s Agentic Opportunity: Moving from breadth to depth with AI agents, defines agent use as activity by Codex and ChatGPT Work. Codex produced 34% of output tokens in the accounts studied, while ChatGPT Work produced 21%. The remaining 45% came from standard ChatGPT use. ChatGPT Work, an agent mode that can work with files, browsers and company tools, launched in July 2026.

Among businesses with 200 or fewer users, weekly active users of Codex or ChatGPT Work rose 8.1 times between the first week of May and the week ending Oct. 5, OpenAI said. The report gave no absolute user counts. It recorded at least fivefold growth in each EU member state, with Belgium at 12.1 times, Italy at 10.8 times and Luxembourg at 10 times. These are changes from the report’s May baseline, not estimates of how many businesses use agents.

The report also found wide differences in agentic output share across countries: Lithuania was at 85% and Cyprus at 81%, compared with Slovenia at 22% and Austria at 26%. Germany was at 51%, ranking 17th among the 27 EU states. OpenAI said its top 10% of firms by output tokens per active user produced about four times as many tokens per user as its typical EU firms. Its non-EU frontier firms produced more than twice the output per user of the EU frontier group.

What the Token Share Can Show

The figures suggest that agent features account for a large portion of generated text in the business accounts OpenAI studied. That matters for companies tracking how employees use AI tools and for policymakers focused on whether firms are moving from basic experimentation toward more involved uses, such as coding or working across company files and applications.

But output tokens are not a direct measure of adoption or economic impact. OpenAI’s Enterprise Signals materials call tokens an imperfect measure of business value and note that agentic tasks can require substantially more computation. An agent may read files, plan, generate a response and revise it, using far more tokens than a short chat exchange. A 55% share of tokens cannot be read as 55% of tasks, users or value.

The business implication is also not settled by growth rates alone. A rise from a small starting base can produce a large multiplier, and OpenAI says a major part of the small-business increase followed ChatGPT Work’s launch. The report points to increased use of its products; it does not establish that agents have already changed productivity across the EU economy.

A Product Launch Shapes the Baseline

The report’s measurement window matters. OpenAI’s chart shows Codex at roughly one-fifth of EU output tokens in May, rising to 34% by Sept. 30. ChatGPT Work was introduced in July and accounted for 21 percentage points by the end of September. That launch contributed more than half of the roughly 35-point increase in agentic share described in the source analysis.

The small-business comparison also starts in early May, before ChatGPT Work existed. OpenAI’s report says a major part of the 8.1-fold rise followed the new mode’s introduction. The growth therefore reflects both increased use of Codex and the availability of another product counted as agentic; the published figures do not separate those effects or give the underlying number of users.

OpenAI’s policy message is that Europe should turn broad AI adoption into deeper use, through worker skills and confidence and wider access to advanced capabilities. Its report estimates that EU firms make up about 10% of frontier firms worldwide, compared with about 65% in North America. Those figures use OpenAI’s account activity and its definition of frontier firms: the top 10% by output tokens per active user.

What the Data Does Not Establish

The report does not publish account or user sample sizes by country, making it hard to judge how representative the national figures are. OpenAI notes that smaller states such as Cyprus, Estonia and Lithuania have relatively high agentic use within larger firms. A small number of large, agent-heavy customers could have a strong effect on a country’s share; the available data does not show how much.

The country rankings also do not provide a simple picture of national readiness. The source analysis found little relationship between agentic output share and small-business growth: Slovenia had the lowest share but above-median growth, while Lithuania had the highest share but below-median growth. These comparisons describe OpenAI’s account base, not all companies in those countries.

More broadly, the report does not show the number of agentic tasks, the proportion of European firms using these products, resulting productivity gains, cost savings or business outcomes. Its findings cover OpenAI’s paid business accounts, not the full EU market. The extent to which the patterns apply to other AI providers or European businesses generally remains unknown.

The Next Measure Is Business Impact

The report sets out a direction for policy and business: improve worker skills and confidence and make advanced AI capabilities more accessible and affordable. It does not announce a new EU program, a regulatory decision or a timetable for further action. The next evidence needed is whether firms move from trying agents to using them routinely, and whether that use produces measurable results.

OpenAI’s published figures provide a baseline for its own products through late September and early October 2026. Further reporting will depend on updates to those usage measures and on whether the company releases absolute user counts, sample sizes and outcome data. Until then, the headline is best understood as a description of where OpenAI’s EU business-account output tokens went—not a market-wide measure of agent adoption or proof of economic gains.

Key Questions

What does OpenAI mean by saying agents generated 55% of output?

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

How much did small-business agent use grow?

OpenAI reported an 8.1-fold rise 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, 2026. It did not publish the absolute user count, and the comparison begins before ChatGPT Work launched.

Which EU countries had the highest and lowest agentic shares?

OpenAI reported the highest share in Lithuania at 85% and the lowest in Slovenia at 22%. These figures reflect output tokens in the company’s accounts, and country-level sample sizes were not disclosed.

Does the report prove agents are improving productivity?

No. The report describes product usage and output-token patterns. It does not establish productivity gains, cost savings or other business outcomes, and OpenAI cautions that tokens are an imperfect measure of business value.

Source: ThorstenMeyerAI.com