Prolepsia forecastYes / no

Will more than 30% of EU enterprises use AI in 2027, as reported by Eurostat?

As of 18 November 2026, Prolepsia puts the chance of yes at 98%.

98%chance of yes

Yes 98%No 2%
Prolepsia forecast made

How Prolepsia reasons

The report Prolepsia wrote with this forecast.

How forecasts are made · Track record

TL;DR

Probability of Yes: 97.7%; Probability of No: 2.3%. The forecast is that Eurostat’s EU-wide figure will be strictly above 30.0% for reference year 2027, with early national surveys providing the strongest evidence.

Key evidence

Eurostat’s official series puts EU enterprise AI use at 8.06% in 2023, 13.48% in 2024 and 19.95% in 2025. The EU needs more than 10.05 percentage points of growth over two survey waves. The gains already recorded from 2023 to 2025 total 11.89 points, so the threshold does not require adoption to accelerate from that pace.

Early 2026 national surveys point to faster growth. Countries accounting for about 28% of eligible EU enterprises recorded a weighted increase of roughly 13 points. Spain rose from 21.1% to 33.6%, with double-digit gains in industry, construction and services. Czechia rose from about 18% to 35%. These are national readings, not a harmonised EU result, but they show that rapid growth was not confined to countries already leading in AI use.

Uncertainties

The early reporters do not establish the pace in Germany, Italy or France. Bitkom’s German survey went from 36% to 57%, but it covers firms with at least 20 employees and its headline gain is not a like-for-like Eurostat measure. The Bank of Italy’s separate survey of firms with at least 20 employees rose from 27% to 32%. These results give reason to expect less growth in some large economies than in the early national releases.

The main path to No is slower growth in those economies followed by a broad slowdown in 2027. The harmonised 2026 EU release is the best next check. Resolution still depends on the first qualifying official figure for reference year 2027, not on the date it is published. Exactly 30.0% at the greatest available official precision does not qualify.

Read the full report

TL;DR

Probability of Yes: 97.7%; Probability of No: 2.3%. The forecast is that Eurostat will report AI use strictly above 30.0% among covered EU enterprises for reference year 2027. The strongest evidence is the jump in early 2026 national surveys; the next figure to watch is the harmonised 2026 EU aggregate.

Context

The question concerns the share of enterprises that use at least one qualifying AI technology for business. Eurostat’s enterprise ICT survey covers firms with at least 10 employees or self-employed persons in specified industries. It counts enterprises, not workers, and does not count firms merely planning to adopt AI. The latest verified EU figure as of November 18, 2026 was still the 2025 observation.

The first complete, non-provisional and comparable EU aggregate for reference year 2027 determines the outcome. The official database takes precedence over a rounded press release. Exactly 30.0% at the highest published precision is No. The forecast is conditional on a usable release; if no usable figure is published by December 31, 2029, the question is annulled rather than resolved No.

Evidence

The official Eurostat series, in the June 15, 2026 database vintage, gives the full observed history: 7.65% in 2021, 8.06% in 2023, 13.48% in 2024 and 19.95% in 2025. There is no 2022 observation and there was no EU-wide 2026 observation at the forecast cutoff. The two most recent annual gains were 5.42 and 6.47 percentage points. Together they add up to 11.89 points, more than the 10.05 points the EU must exceed over the next two survey waves. That is a useful benchmark, not a dependable trend by itself. There are only two recent annual changes, and the 2025 questionnaire added a category for generating images, video or audio.

Paid cloud services offer a check on the idea that enterprise technology adoption keeps rising at the same speed. The comparable Eurostat cloud series went from 23.90% in 2018 to 36.10% in 2020, then from 45.32% in 2023 to 52.74% in 2025. The gain in the latter two-year span was smaller than in the former. Cloud is not a model for AI adoption. It shows why a strong survey wave need not be followed by another one just as strong.

The early 2026 national readings carry more weight than a straight-line extrapolation. Across ten early reporters representing about 28% of eligible EU enterprises on proxy enterprise counts, adoption rose roughly 13 percentage points on a weighted basis. Spain rose from 21.1% to 33.6% in its national survey, and Czechia rose from about 18% to 35%. The gains extended to countries starting farther below the EU average: Poland rose from 8.7% to 17.8%, and Latvia rose from about 12.2% to 18.5%. The pace varied, but the direction did not depend on a single country or on countries that already had high adoption.

Germany and Italy provide weaker, but consequential, evidence. Bitkom’s September survey found that use among German firms with at least 20 employees rose from 36% in 2025 to 57% in 2026. Scaling that change by the prior year’s relationship between official and Bitkom growth implies about eight points, not 21. That is a rough inference, not an official German reading. The Bank of Italy’s separate survey of firms with at least 20 employees rose from 27% to 32%. Both surveys show growth, but neither has the same population as Eurostat. Taken together, they temper the inference that the largest economies matched the early reporters.

What's non-obvious

The jump is not just a story about which industries were counted. Spain changed its industry classification and added activities, so its headlines are not a clean like-for-like comparison. But its national releases still show gains of 11.6 points in industry, 11.2 in construction and 12.5 in services (2025 survey; 2026 survey). A shift in the mix of those broad sectors cannot explain most of the jump. Sweden also widened its coverage, but adoption among firms with 10–49 employees rose from 30.8% to 45%, and 87% of 2026 adopters reported using off-the-shelf closed-source tools (Statistics Sweden). Firms do not need to build an AI system to count as users.

That matters because adoption among smaller firms has more room to move the EU total. In 2025, around 83% of eligible enterprises had 10–49 people employed. Across all firms with 10–249 people employed, AI use was 18.9%, against about 55.0% among large firms (Eurostat). The large-firm figure shows that AI use was already common in one part of the economy. The small-firm evidence is more informative about how much farther the enterprise-count aggregate can rise.

Uncertainties

The national releases are early, not representative. The roughly 28% coverage estimate uses structural business counts as a proxy for the ICT survey frame, and the early reporters grew faster than the rest of the EU in both recent years used for comparison. Survey timing, industry coverage and rounding also differ. Germany, Italy and France together make up about 48% of the proxy enterprise count, and comparable 2026 readings for those countries were not available at the cutoff. The gap in published data is not evidence that their adoption stalled; it is why the early national results cannot establish the EU-wide gain.

The clearest route to No is a combination: growth in the large unreported economies falls short of what the early releases suggest, then adoption slows sharply in the 2027 survey wave. Measurement can also matter near the line. The 2027 regulation retains the eight core AI-use questions but uses a revised industry classification. The harmonised 2026 release will narrow the largest uncertainty by showing how far the EU actually moved from 19.95% before the final survey wave. Only the qualifying 2027 figure settles the question.

Forecast history

How Prolepsia’s forecast has movedOne forecast so far: 98% on 18 Nov 2026.
DateForecast
98% chance of yes

Sources Prolepsia read

  1. ICT usage in enterprises (isoc_e)ec.europa.eu
  2. Nota de Prensa: Encuesta sobre el uso de TIC y del comercio electrónico en las empresas. Año 2025 - Primer trimestre 2026. Datos definitivos.www.ine.es
  3. AI používají v Česku tři čtvrtiny velkých firem | Produktycsu.gov.cz
  4. pdf https://publikacje.new.stat.gov.pl/sites/default/files/2026-10/information_society_in_poland_in_2026.pdf Information society in Poland in 2026. News release in PDF format 509.16KB (Kilobytes)publikacje.new.stat.gov.pl
  5. More enterprises use artificial intelligence | Oficiālās statistikas portālsstat.gov.lv
  6. Erstmals nutzt die Mehrheit der Unternehmen KI | Presseinformation | Bitkom e. V.www.bitkom.org
  7. Almost every second enterprise used AI in 2026www.scb.se
  8. Implementing regulation - EU - 2026/1566 - EN - EUR-Lexeur-lex.europa.eu
  9. It i virksomheder - Danmarks Statistikwww.dst.dk
  10. 65% of people and 34% of enterprises use AI | Statistikaametstat.ee

The question

This question asks whether more than 30% of enterprises in the European Union (EU) will use artificial intelligence (AI) technologies in 2027, according to Eurostat's annual survey on ICT usage and e-commerce in enterprises. The relevant indicator measures the percentage of enterprises using at least one AI technology to conduct their business activities. Following Eurostat's standard methodology, the population covers enterprises with at least 10 employees or self-employed persons in the economic activities included in the survey. The measure encompasses small, medium-sized, and large enterprises across EU member states, rather than all businesses regardless of size. AI technologies include machine learning, natural language processing, text analysis, speech recognition, image recognition, generative AI, and other AI-based systems included in Eurostat's survey definition. As of November 18, 2026, the latest verified annual figures show that 19.95% of EU enterprises used AI in 2025, compared with approximately 13.5% in 2024 and 8.0% in 2023. This represents an increase of approximately 6.5 percentage points between 2024 and 2025, following an increase of approximately 5.5 percentage points between 2023 and 2024. AI adoption varies substantially across countries and enterprise sizes. In 2025, 55.0% of large enterprises used AI, compared with 18.9% of small and medium-sized enterprises. Continued adoption could bring the EU-wide percentage above 30% by 2027, although the pace of adoption remains uncertain. The question concerns the statistical reference year 2027, not the percentage reported or published during calendar year 2027. The outcome will be determined once Eurostat publishes the relevant 2027 data, expected after the reference year. Background sources: Eurostat, 'Use of artificial intelligence in enterprises' (https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Use_of_artificial_intelligence_in_enterprises); Eurostat, '20% of EU enterprises use AI technologies,' December 11, 2025 (https://ec.europa.eu/eurostat/web/products-eurostat-news/w/ddn-20251211-2).

How it resolves

The question resolves YES if Eurostat's published statistics for reference year 2027 show that strictly more than 30.0% of EU enterprises used at least one AI technology to conduct their business activities. It resolves NO if the corresponding percentage is less than or equal to 30.0%. The primary resolution source will be Eurostat's dataset 'Enterprises using AI technologies' (online data code: isoc_eb_ai), using the EU aggregate for all enterprise size classes covered by the survey, the total of the included economic activities, and the indicator representing enterprises using at least one AI technology. The observation must refer to 2027, regardless of the publication date. The percentage will be evaluated using the greatest numerical precision available in Eurostat's published data. A reported value of exactly 30.0% will resolve NO unless more precise official data establish that the underlying percentage exceeds 30.0%. The first complete, officially published EU aggregate for reference year 2027 will determine the outcome, provided that it is not explicitly marked as provisional or subject to an announced correction. If the initial release is provisional, the question will be resolved using the first subsequent non-provisional release. Later routine revisions will not change the resolution. If Eurostat renames, replaces, or reorganizes the dataset, an official successor indicator measuring the same underlying concept will be used. If the methodology changes, the official 2027 EU-wide figure will be used as long as it remains substantively comparable to the historical indicator. If Eurostat discontinues the indicator without publishing a comparable replacement, or if no usable EU-wide figure for 2027 has been published by December 31, 2029, the question will be annulled rather than resolved YES or NO. Eurostat's official statistical database will take precedence over rounded figures in press releases, secondary reporting, or unofficial estimates.

Fine print

The 30% threshold is strict: exactly 30% does not qualify. The denominator consists of enterprises covered by Eurostat's ICT usage and e-commerce survey, rather than every registered enterprise in the EU. Enterprises with fewer than 10 employees or self-employed persons are generally excluded. The measure is based on the proportion of enterprises reporting AI use, not the proportion of employees using AI, the proportion of enterprises developing AI, or the proportion investing in AI. An enterprise qualifies if it uses at least one AI technology covered by Eurostat's applicable survey methodology, regardless of whether the technology was developed internally or obtained from an external provider. Individual EU member states exceeding 30% will not independently satisfy the question; only the overall EU aggregate counts. The geographic scope is the European Union as defined by Eurostat for the 2027 reference year, excluding non-EU countries. If EU membership changes before or during 2027, the official Eurostat EU aggregate for that reference year will be used. The outcome is based on reported adoption during the 2027 statistical reference period, not on whether the threshold was temporarily exceeded at some point during the year. Changes in the list of qualifying AI technologies will follow Eurostat's official methodology for the 2027 survey, provided the resulting indicator remains substantively comparable to the historical series.