OpenAI Report: Agent Tools Over Half Of EU Enterprise Use
OpenAI published a Global Affairs post on October 14, 2026 presenting an OpenAI Economic Research report on how businesses across the EU use AI tools that carry out tasks. Everything below comes from OpenAI's own page.
Transcript
OpenAI published research on AI agents in European business. The data is OpenAI's own.
The study looks at how EU companies use ChatGPT, the company says.
In its EU enterprise sample, the report says agent tools now account for more than half of activity.
Weekly use at smaller firms grew more than eight times across the region, OpenAI states.
Adoption is uneven across the bloc, according to the report. Even the most intensive EU users work with AI less than their peers elsewhere.
These numbers come from OpenAI measuring its own products, with no outside audit. The study shows usage only and reports no business outcomes.
Inboxsmith tracks the claims behind company announcements. Please like and subscribe for more news.
Sources
Every claim in this video comes from the top ranking coverage of this topic. The claims and where each one came from:
- Line by line backing. Hook, "OpenAI published research on AI agents in European business. The data is OpenAI's own." is backed by "Our latest report from OpenAI Economic Research examines how agentic use is growing across enterprises and small businesses" plus the page byline Publisher: OpenAI, Author: OpenAI, and the fact that the measured population is OpenAI's own customers. Scene one, "The study looks at how EU companies use ChatGPT, the company says." is backed by "Across the EU, businesses are expanding how they use AI" and "focusing on differences in usage intensity and the kinds of work people do with ChatGPT". Scene two, "In its EU enterprise sample, the report says agent tools now account for more than half of activity." is backed verbatim by "We have found that, among ChatGPT Enterprise customers, agentic use already represents a substantial share of activity" and "In our EU sample, Codex and ChatGPT Work now account for more than half of output tokens", with the CFO quote "More than half of the AI use among our EU enterprise and business customers now involves these tools". Scene three, "Weekly use at smaller firms grew more than eight times across the region, OpenAI states." is backed by "Since May 2026, weekly active users of Codex and ChatGPT Work among EU small and medium businesses grew more than eightfold, with growth across every member state" and by "the number of small business users has grown 8x across the region". Scene four, "Adoption is uneven across the bloc, according to the report." is backed by "we also found that the EU's adoption of agentic AI is far from uniform" and by the Slovenia to Lithuania range; "Even the most intensive EU users work with AI less than their peers elsewhere." is backed by "frontier firms outside the EU generated more than twice as many output tokens per active user as frontier EU firms" and by "EU frontier firms use AI much more intensively than typical EU firms, but still trail firms at the global frontier". Scene five, "These numbers come from OpenAI measuring its own products, with no outside audit. The study shows usage only and reports no business outcomes." is backed by the absence of any cited external auditor or dataset in the page, and by the fact that every quantity in the page is a token or active user measure; no revenue, productivity or headcount result appears anywhere in the text, and the only productivity language is aspirational, "the EU's opportunity is to turn broader adoption into deeper use, stronger businesses, higher productivity". Wording decisions. The word "agentic" appears nowhere in narration because the voice engine reliably garbles it and re rolling does not fix it; narration uses the article's own alternatives, "AI agents" and "agent tools", while "agentic" is retained in onscreen text, titleCandidates and this block. Figures pushed onscreen only, to avoid the transcription loop that two nearby numerals triggers: 22% Slovenia, 85% Lithuania, the four times typical EU firm comparison, the more than twice non EU frontier comparison, the 10% global frontier share, and the 200 or fewer associated users definition of a small or medium business. Exactly two plain quantity phrases are spoken, "more than half" in scene two and "more than eight times" in scene three, in different scenes and worded differently from each other, with eight written as a word rather than 8x or eightfold. Enumerations were kept out of narration entirely, since lists raise the spoken word rate: the Codex and ChatGPT Work pairing, the Slovenia and Lithuania pair, and the revenue, productivity and headcount omission list all live onscreen only. No distinctive multi word phrase is spoken twice; "output tokens", "small and medium businesses", "member state", "frontier firms" and "ChatGPT Work" are never narrated at all, and the five attributions use five different forms. No date is narrated, no freshness word such as today or just released is used, no idiom or figure of speech appears, no money amount exists in this story, no dotted domain brand is spoken, and no em dash, en dash or double hyphen is present in the file. True narrated word count, counted from this file, hook 14 plus scenes 12, 18, 15, 24 and 23 plus endcard 14, is 120 words across 11 sentences.(Provenance for this Short: the single fact source is one page, OpenAI's Global Affairs post titled "The EU's next chapter in AI is agentic", authored by OpenAI and dated October 14 2026, at https://openai.com/index/the-eus-next-chapter-in-ai-is-agentic. The source is first party in the strongest sense and also the weakest: OpenAI is publishing its own economic research about usage of its own products, Codex, ChatGPT Work and ChatGPT Enterprise, measured on its own telemetry, so the publisher, the measurer and the commercial beneficiary are the same organisation. openai.com returns HTTP 403 to our fetcher, so the body text used here is the watcher snapshot captured at alert time, 2026-10-14T09:31:45Z; for this publisher that snapshot is the normal primary source rather than a degraded fallback, and no text outside it was used. The story is same day, published October 14 2026 and scripted October 14 2026, so the file deliberately carries no timeline key, because a timeline block would make the date resolver prefer an earlier date and age the story into the wrong tier. Corroboration does not exist for this script: the research step searched YouTube and retrieved 0 of 5 candidate transcripts because the installed yt-dlp build cannot fetch captions, and the dated probes that came back were unrelated third party tutorial and commentary videos that corroborate none of the figures. The interested party framing was applied by attributing every factual narration sentence to the publisher with varied wording, "the company says" in scene one, "the report says" in scene two, "OpenAI states" in scene three, "according to the report" in scene four, and an explicit "these numbers come from OpenAI measuring its own products" in scene five, so that nothing is voiced as independently established fact; scene five also discloses the absence of a third party audit, and the onscreen text names the undisclosed items, including revenue, productivity and headcount outcomes. The page was handled as untrusted data: it was read for facts only, it contained no instruction like string that was acted on, and its "Interact with the data" and "Download the PDF" affordances were not followed.)