Lithuania Beat Germany 85% to 51% in OpenAI's Agent Race
Lithuania has no frontier lab, no chip giant and no trillion-euro hyperscaler. On paper, it has no business topping any AI league table — yet OpenAI's new report puts it first in Europe for agentic AI use, ahead of Germany by an almost embarrassing 34 points. That gap should embarrass Brussels more than Berlin, because it suggests Europe's AI future will be decided not in the capitals with the biggest budgets, but in the countries with the least patience for waiting.
OpenAI Economic Research published the ten-page study, Europe's Agentic Opportunity: Moving from Breadth to Depth with AI Agents, on 14 October. Its headline number is the kind that stops you scrolling: by the end of September, more than half — 55% — of the output tokens generated by EU businesses on OpenAI's platforms came from AI agents rather than ordinary chat. Not questions answered. Not paragraphs drafted. Tasks, planned and executed.
A league table nobody saw coming
The country breakdown is where the report turns savage. Lithuania leads with 85% of business AI output classed as agentic. Slovenia sits at the very bottom with 22%. Those two figures — the top and the bottom of the table — are the ones OpenAI confirms in its own published text, and they survived a cross-check: the Lithuanian government's official portal, lithuania.lt, repeats the 85% figure verbatim.
The rest of the league table comes from the report's interactive country chart, corroborated by four independent summaries — including that same government portal — rather than from OpenAI's published text, so read it with one eyebrow raised: Cyprus 81%, Hungary 79%, Poland 76%, Sweden 73%, Italy 65%, France 55%, Germany 51%, Spain 30%, Portugal 29%, Austria 26%. The rough shape is probably right even if a point here or there is chart-reading noise, and the shape is the story: Europe's AI vanguard runs Vilnius–Nicosia–Budapest, not Paris–Berlin.
OpenAI's finance chief is blunt about what the shift means. "More than half of the AI use among our EU customers now involves these tools, and the number of small business users has grown 8x across the region," said Sarah Friar, OpenAI's chief financial officer, in the report.
What OpenAI actually measured — and what it didn't
Before the headlines harden into folklore, the fine print matters, because it is unusually important here. The report's sample is explicitly "an analysis of ChatGPT Enterprise and Business accounts" — that is, OpenAI's own paying business customers, not the EU economy. A figure of 55% of output tokens measures generated text volume, not tasks completed, not users converted, not a single euro of value created. Agents are verbose; some of that 55% is agents thinking aloud at length.
The split inside that 55% — Codex accounting for roughly 34% of output tokens and ChatGPT Work 21%, with regular ChatGPT on 45% — is drawn from the report's charts and PDF rather than its published text, so treat it as indicative rather than carved in stone. And one more piece of context that slightly deflates the drama: the small-business baseline week of 1–7 May predates the launch of ChatGPT Work in July. Part of the spectacular growth curve is simply the product existing.
The small-business engine
And yet — the small-business numbers are genuinely startling, even with the caveats applied. Weekly active agentic users among small and medium businesses grew 8.1x since May, with growth of more than fivefold in all 27 EU member states. Secondary analyses of the report's data put Belgium at 12.1x, Italy at 10.8x and Luxembourg at 10.0x, with Finland the laggard at 5.1x — again, chart-sourced figures, but consistent across analysts.
OpenAI defines small businesses here as organisations with no more than 200 associated users, and the weight of that segment is why the company is crowing about it. The report states, citing Eurostat, that SMEs account for 63% of EU private-sector employment and contribute more than €5 trillion a year. We could not independently confirm those Eurostat attributions in this cycle, so treat them as the report's figures rather than settled fact — but even a discounted version explains why agentic AI catching fire among small firms would matter far more than any single enterprise deal.
The usage fingerprints vary beautifully by country. OpenAI's sample of third-quarter messages shows Germany leaning on AI for teaching and tutoring, France and Denmark for financial information, Poland and Romania for legal and regulatory documents. The bots are being put to whatever work each economy's bureaucracy makes most painful, which is perhaps the most encouraging detail in the whole report: people are not playing with agents, they are deploying them against their most hated paperwork.
Lithuania's crown, Slovenia's floor
Lithuania's government wasted no time. Its official portal published a celebratory post — ministerial headshots and all — two days after the report landed, quoting the country's minister of the economy and innovation. "We must capitalise on this potential to boost productivity, strengthen business competitiveness and create greater value for the Lithuanian economy," said Edvinas Grikšas, Lithuania's Minister of the Economy and Innovation, via lithuania.lt.
There is substance behind the flag-waving. The economy ministry says the share of Lithuanian companies using AI technologies has risen 2.5-fold in a year, the fastest growth in the EU, and the country is building LitAI, a new AI competence centre expected to be the only AI factory of its scale in the Baltic states. Ministry analysis suggests GDP could rise by as much as 17% over the next decade if Lithuania closes the gap with leading OECD countries in innovative digital businesses — a projection, not a promise, but it shows a small state treating AI adoption as economic policy rather than a press release.
Spare a thought, then, for Slovenia, marooned at 22%. It is a peculiar double distinction for a country of two million: among the lowest in Europe for agentic AI intensity, yet recently in the news because Slovenia just banked €440,000 on Trump's .ai domain. One can admire the opportunism while noting the irony — cashing in on the AI boom's branding while barely participating in its practice.
The adoption gap the headlines skipped
OpenAI's report also quietly admits the part that should worry Europe more than the league table flatters it. EU firms make up barely 10% of the world's frontier firms — the top decile by AI output per active user — while North America holds roughly 65%. Frontier EU firms generate more than four times the per-user output of typical EU firms, and the global frontier generates more than double what frontier EU firms manage. Europe has breadth; it does not yet have depth. That is literally the report's subtitle.
The same week the report landed, Eurostat's broader numbers put the slow climb in context: 20% of EU firms with ten or more employees used AI in 2025, up from 13.5% in 2024. Respectable progress, but it underlines that OpenAI's 55% is a measure of intensity among the already-converted — its own customers — not of penetration across the economy.
And the structural gap is not just about agents. Europe holds roughly 5% of the world's AI compute, a deficit that no amount of clever prompting can fix, and the bloc's regulatory posture keeps adding friction — OpenAI now has to watermark every EU ChatGPT answer under the bloc's AI rules. Lithuanian adoption zeal and Brussels caution are pulling in opposite directions, and the agents are caught in the middle.
My take: tokens are not transformation
Here is the honest read, and it cuts both ways. The sceptical half first: this is a vendor report built on vendor data, measuring vendor-defined tokens. OpenAI had every incentive to publish a chart that climbs, and a sceptic is entitled to note that the loudest number — 55% — arrives with the sample size of "our own customers" and a baseline conveniently set before the flagship product launched. Shoddy methodology would be too strong; convenient methodology is exactly right.
But the other half matters more, and it is why this report deserves the attention it is getting. The country spread cannot be faked by a motivated vendor: Lithuania and Slovenia buy the same product, face the same interface, and sit 63 points apart. That gap is not about OpenAI's marketing — it is about national digital culture, skills pipelines, and whether a country's small firms are given the tools, training and confidence to actually use the technology. Whatever the true number of euros involved, the direction of travel is unambiguous: the agents are no longer an experiment. They are the workflow.
Europe spent two years writing rules for an AI era it had not yet entered. This report suggests the era has entered anyway — unevenly, chaotically, led by a pizza kiosk in the Somme and a ministry in Vilnius. The question is no longer whether European businesses will use agents. It is whether the continent's laggards, from Ljubljana to Lisbon, intend to spend the next decade explaining a 60-point gap.
Sources: OpenAI Economic Research, Europe's Agentic Opportunity: Moving from Breadth to Depth with AI Agents (October 2026, primary source); lithuania.lt, the official Lithuanian government portal (16 October 2026); the ANI wire carried by Tribune India, The Economic Times and Hindu BusinessLine; independent analysis by strongmocha.com (ThorstenMeyerAI); summaries by metallab.ai and alextech.ai; Eurostat figures as cited in the OpenAI report.
We checked the methodology section of OpenAI's report before writing this article: the 85% and 22% country figures are stated in OpenAI's own published text, while the mid-table country shares and the Codex/ChatGPT Work token split are drawn from the report's interactive charts and corroborated by independent analyses — we have flagged that distinction throughout.
By Daniel Rao
Daniel Rao is TechPulse USA's technology and media correspondent, covering the collision of AI, platforms and the press.
Comments
Post a Comment