Skip to content

Reviewed edition · qasb report 10/2026 · Topic report · Organizations

AI in organizations: more businesses use AI

In several countries more businesses used AI in 2026 than a year earlier, large firms far more often than small ones. Many users spend little or nothing on it, and evidence on effects rests mostly on self-reports.

Published
October 27, 2026 · as of October 26
Reading time
12 min
Publisher
qasb Research
Method
qasb methodology v1

Key findings

  1. 24.6%of U.S. businesses reported using AI in the previous two weeks, up from 17.3% in the survey of 17–30 November 2025.1Evidence: observed
  2. about 30%of covered UK businesses used at least one AI technology in September 2026, up from 9% in 2023.2,12Evidence: observed
  3. 84.7%of Austrian enterprises with 250 or more employees used AI in 2026, against 44.0% of those with 10 to 49 employees.7Evidence: observed
  4. 50%of AI users among French businesses with up to 249 employees used only free solutions in 2026.13Evidence: observed

1.0Summary

In Denmark the share of businesses with ten or more employees using AI rose from 15% (2023) to 59% (2026), in Austria from 30% (2025) to 48.1%. The official surveys show diffusion, not effects, and are not directly comparable across countries.

  • Large firms use AI far more often than small ones, in Austria 84.7% of those with 250 or more employees versus 44.0% of those with 10 to 49. By sector, information and IT lead.
  • German establishments neither using nor planning AI in 2025 mainly cited missing use cases (almost 90%). Every other reason was named by only about a quarter to just over a third.
  • Many users spend little or nothing on AI. In France, half of AI users among small and medium-sized businesses relied only on free solutions.
  • Effects on productivity and employment rest mostly on self-reports. Only one older study, on a narrow use case, rests on an experiment.

2.0Adoption: shares keep rising

Official surveys from seven countries show how many businesses use AI or plan to use it. In the United States, Austria, Denmark and Norway the share rose further in 2026, and in the United Kingdom it is well above its level when the question was introduced in 2023. For Canada, this edition shows planned use, and for Spain there is no comparable previous-year value.

Figure 1

AI use by Danish enterprises

Share of enterprises with ten or more employees using AI (Statistics Denmark)

Private non-agricultural activities excluding mining and financial and insurance activities, table without uncertainty measures. This edition does not include a value for 2024. The figures measure use of at least one AI technology, not intensity or effects.Evidence: observed · Sources 16

In the U.S. Census Bureau's BTOS file of 22 October 2026, 24.6% of U.S. businesses reported using AI in at least one business function in the two weeks before the survey; at the start of this question series (collection 17–30 November 2025) the share was 17.3%.1A1

According to the ONS BICS survey, around 30% of covered UK businesses used at least one AI technology in September 2026, and 38% of businesses with ten or more employees; when the question was introduced in September 2023, the overall share was 9%.2,12A2

According to Statistics Denmark, 59% of Danish enterprises with at least ten employees used AI in 2026, up from 15% in 2023 and 42% in 2025; the share was 55% among firms with 10 to 49 employees and 88% among those with 250 or more.16A3

Further findings

  • In Austria, 48.1% of enterprises with ten or more employees used AI technologies in 2026, up from 30% in 2025.7A4
  • In Norway, 48% of enterprises with ten or more employees used AI in 2026, up from about three in ten in 2025 (new industry classes).19,21A5
  • In Spain, 33.6% of enterprises with ten or more employees used AI in the first quarter of 2026, not strictly comparable with earlier years.4A6
1 more finding
  • In Canada, 25.2% of businesses planned in Q3 2026 to use AI for goods or services, and 52.7% did not.3A7

3.0Differences: size, sector, area of use

Use differs strongly by firm size and sector. In the United States, large firms and the information sector and professional, scientific and technical services lead, and the share in finance and insurance is also high. In Austria, users most often name marketing or sales as their area of use.

Figure 2

AI use by firm size in the United States

Share of businesses using AI in the last two weeks, BTOS file of 22 October 2026

Self-reports without information on intensity or effects. Mid-sized classes have larger standard errors (up to 1.97 percentage points). At the start of the series (collection 17–30 November 2025) the values were 17.0% (1 to 4) and 30.5% (250+).Evidence: observed · Sources 8

In the BTOS file of 22 October 2026, 43.6% of U.S. businesses with 250 or more employees reported using AI in the last two weeks, compared with 24.3% of businesses with 1 to 4 and 22.6% of those with 5 to 9 employees.8A8

By sector, in the BTOS file of 22 October 2026, 40.5% of U.S. businesses in finance and insurance, 27.6% in health care and social assistance, 22.9% in manufacturing and 18.5% in retail trade reported using AI in the last two weeks.9A9

Further findings

  • Austrian enterprises using AI most often used it in marketing or sales in 2026 (67.5%), and only 9.2% in logistics.7A10

4.0Barriers: missing use cases and expertise, cost

In Germany, establishments without AI mainly cite missing use cases, while the EU sources often name a lack of expertise, cost and legal uncertainty. Each source covers a different group, and the EU survey also includes automated systems without AI and refers to workforce-related uses. The analysis by the EU's Joint Research Centre (JRC) describes relative shifts, not absolute rates.

Figure 3

Reasons for not using AI in German establishments

Establishments neither using nor planning AI, second half of 2025, multiple answers

Self-reports without uncertainty measures. Only establishments without use or plans were asked, about two thirds of all establishments. The source gives the first value as 'almost 90%'.Evidence: observed · Sources 10

According to the IAB, establishments in Germany that neither used nor planned to use AI in the second half of 2025 cited the lack of use cases by far most often (almost 90%); every other reason was named by only about a quarter to just over a third, including missing knowledge (36%), a lack of skilled staff for implementation (34%), incompatibility with existing hardware and software (33%), data protection and security concerns (32%), legal uncertainty (32%) and excessive cost (30%).10A11

In Flash Eurobarometer 578, 50% of surveyed EU companies named high investment costs, 49% legal or regulatory uncertainty and 48% a lack of internal expertise as an important difficulty for using AI or automated systems in workforce-related areas.5A12

According to the JRC's analysis of the Eurostat ICT survey, a lack of relevant expertise remained the main barrier, cited by roughly 70% of firms that had considered but not adopted AI; regulatory uncertainty (data protection, legal consequences) increased most between 2023 and 2025, while concerns about cost, data availability and compatibility declined.6A13

5.0Implementation: small budgets, training, rules, oversight

Four sources ask about AI spending, and only the UK survey gives amounts, in bands. They show small or missing budgets, much uncertainty about spending and often no training. Among EU companies using AI or automated systems at the workplace, the final decision mostly rests with people, by their own account. For financial institutions, the Financial Stability Board (FSB) published non-binding practices.

Figure 4

Investment in adopting AI in the United Kingdom

Businesses with ten or more employees using AI or unsure whether they do, investment over twelve months

ONS BICS survey, wave 165, weighted shares. Self-reports in value bands without exact amounts; the high 'not sure' share limits interpretation, and among businesses with 250 or more employees 57.4% were not sure. Finance and insurance are not covered.Evidence: observed · Sources 12

Among UK businesses with ten or more employees that used AI or were unsure, 35.0% reported investing less than £10,000 in adopting AI over the last twelve months, 7.8% between £10,000 and £49,999 and 1.1% £500,000 or more; 23.4% had not invested and 28.1% were not sure, according to the ONS.12A14

According to the France Num Barometer 2026, 40% of French businesses with 0 to 249 employees used AI solutions (2024: 13%), micro-enterprises (TPE) 38% and PME 53%; among AI users, 50% used only free solutions and 43% had spending on subscriptions or licences.13A15

According to the ONS, 33.9% of UK businesses with ten or more employees that used AI or were unsure had not trained any staff in using AI; 25.5% had trained 1 to 24% of their workforce, 6.6% more than three quarters, and 22.3% were not sure.12A16

Among companies in Eurobarometer survey 578 that used AI or automated systems at the workplace, only 2% reported fully automated decisions without human intervention; 52% said the systems make suggestions and humans always decide, and 38% said decisions are made entirely by humans.5A17

Further findings

  • According to an older IAB study, 45% of German establishments using generative AI had spent money on it, and 21% had internal rules.14A18
  • In a non-representative survey of non-profit organisations, 57% of executives reported having no dedicated AI budget.17A19
  • On 15 October 2026 the Financial Stability Board (FSB) published twelve non-binding practices for AI adoption by financial institutions.11A20

6.0Effects: mostly self-reported

Evidence on whether AI changes productivity or employment so far comes mostly from firms' own statements. The figures from France and China come from surveys and interviews, not from measured productivity data. Only one older study, on a narrow use case, is based on an experiment.

Figure 5

Generative AI in French firms

Banque de France survey of about 7,000 firms, early 2026

The firms' own reports, not measured productivity. The shares refer to different groups: firms with 20 or more employees, users of generative AI and the balance of reports.Evidence: observed · Sources 15

In the Banque de France survey of about 7,000 firms in early 2026, 67% of firms with 20 or more employees used generative AI, mostly experimentally or to a limited extent; only 31% of firms using generative AI reported productivity gains already observed, and on balance 2% reported an AI-related decline in their workforce.15A21

An ILO research brief on 21 Chinese enterprises found that all of them used AI or had concrete plans to adopt it, but most lacked systematic evaluation frameworks; no firm described a controlled evaluation, and the reported productivity gains rest on the firms' own statements.18A22

In the Banque de France survey, only 31% of firms using generative AI reported productivity gains already observed.

Further findings

  • Field experiment with vending machines: capped overrides of AI restocking suggestions cut inventory without a significant loss of sales.20A23

7.0Evidence

Not every statement in this report is equally well supported. The table shows the key statements with their evidence type, AI relationship and sources.

StatementEvidenceAI linkSources
24.6% of U.S. businesses reported using AI in the previous two weeks, up from 17.3% in the survey of 17–30 November 2025observednot stated1
about 30% of covered UK businesses used at least one AI technology in September 2026, up from 9% in 2023observednot stated2,12
84.7% of Austrian enterprises with 250 or more employees used AI in 2026, against 44.0% of those with 10 to 49 employeesobservednot stated7
AI use in the US: 43.6% at 250 or more employees, 22.6% at 5 to 9observednot stated8
German firms without AI mostly cited a lack of use cases (almost 90%)observednot stated10
50% of AI users among French businesses with up to 249 employees used only free solutions in 2026observednot stated13
31% of firms using generative AI in France already saw productivity gainsobservednot stated15
None of 21 Chinese firms described a controlled evaluationobservednot stated18

The 23 reviewed statements from 21 original sources fall into chapters 2.0 (7), 3.0 (3), 4.0 (3), 5.0 (7) and 6.0 (3); all are classed as observed. “Not stated”: the edition does not assess the AI relationship of these statements separately. The scope and limitations of every statement are listed in the appendix (A1 to A23).

Original sources checked on October 26, 2026.

8.0What organisations can do now

  1. Check use cases: establishments in Germany without AI mainly cite a lack of use cases. Statistics such as Austria's show where others often use AI (marketing and sales), but not whether that suits one's own business or what it yields.
  2. Set responsibilities, training and rules: about a third of UK businesses with ten or more employees that use AI or are unsure have trained no staff. For financial institutions, the FSB describes twelve non-binding practices.
  3. Measure effects rather than assume them: plan evaluation early. Reported productivity gains rest mostly on self-reports. In interviews with 21 Chinese firms (ILO research brief), none described a controlled evaluation.
  4. Design override rights: in an older, non-peer-reviewed preprint on a single firm, capped overrides of AI replenishment suggestions cut inventory without a significant sales decline, while free overrides also cut sales. Whether this holds elsewhere is open.

9.0Method and limitations

Reporting period: October 1, 2026 up to the data cutoff of October 26, 2026 at 1:00 PM GMT+1 (Berlin time). This topic report is published before the end of the month; sources published after the cutoff follow in the next edition of the series. Compiled: October 27, 2026 at 12:23 AM GMT+1. This is the first baseline. The sources come from different years; their publication dates are not events in the reporting month. No change from an earlier edition was measured.

The report uses only reviewed public original sources. Customer conversations, private dossiers and personal customer assessments are not inputs. Observed, model estimate, scenario and open describe the evidence for each finding. Source values retain their original units and baselines; they are not combined into an overall score.

Editorial approval: qasb editorial team · Approved: October 27, 2026 at 12:25 AM GMT+1. The editorial presentation arranges the reviewed statements and contains no figure that is not in them. Presentation approved by the qasb editorial team · October 27, 2026 at 2:27 PM GMT+1.

Limits

  • This topic report covers one topic area: “Organizations”. It is not an exhaustive survey of all studies or incidents.
  • A narrow study finding is not a forecast for a particular customer, occupation or security system.
  • No qualified finding does not prove no effect; source-access gaps are disclosed separately.
  • This report contains no illustrative measurement series or invented confidence rating.

10.0Sources

  1. 1Business Trends and Outlook Survey (BTOS): National current core estimates (National.xlsx)U.S. Census Bureau, Business Trends and Outlook SurveyOctober 22, 2026www.census.gov/hfp/btos/downloads/National.xlsx
  2. 2Business insights and impact on the UK economy: 8 October 2026Office for National Statistics (ONS)October 8, 2026www.ons.gov.uk/businessindustryandtrade/business/businessservices/bulletins/businessinsightsandimpactontheukeconomy/8october2026
  3. 3Analysis on planned use of artificial intelligence by businesses in Canada, third quarter of 2026Statistics Canada (B. Li, R. Fair, C. Johnston), Catalogue no. 11-621-MOctober 1, 2026www150.statcan.gc.ca/n1/pub/11-621-m/11-621-m2026013-eng.htm
  4. 4Encuesta sobre el uso de TIC y del comercio electrónico en las empresas. Año 2025 – Primer trimestre 2026. Datos definitivosInstituto Nacional de Estadística (INE), SpainOctober 22, 2026www.ine.es/dyngs/Prensa/ETICCE20251T2026.htm
  5. 5Flash Eurobarometer 578: Workplace-related drivers and obstacles for the adoption of Artificial Intelligence and Algorithmic Management by SMEs and Mid-Caps (processed data, Volume A)European Commission, Directorate-General for Communication (Eurobarometer)October 14, 2026data.europa.eu/data/datasets/s3696_fl578_eng?locale=en
  6. 6Measuring AI adoption in the European Union – Methodology and results for strategic sectors (JRC146753, EUR 40868)European Commission, Joint Research Centre (M. Ignaszak, D. Nepelski, J. Stake, A.-A. André)October 14, 2026doi.org/10.2760/7714638
  7. 7IKT-Einsatz in Unternehmen 2026: Ergebnisse (Tabellenbände ohne und mit Wirtschaftszweig Finanz- und Versicherungsleistungen, erstellt am 20.10.2026)Statistik Austria (Bundesanstalt Statistik Österreich)October 20, 2026www.statistik.at/statistiken/forschung-innovation-digitalisierung/digitale-wirtschaft-und-gesellschaft/ikt-einsatz-in-unternehmen
  8. 8Business Trends and Outlook Survey (BTOS): Employment Size Class current core estimates (Employment Size Class.xlsx)U.S. Census Bureau, Business Trends and Outlook SurveyOctober 22, 2026www.census.gov/hfp/btos/downloads/Employment%20Size%20Class.xlsx
  9. 9Business Trends and Outlook Survey (BTOS): Sector current core estimates (Sector.xlsx)U.S. Census Bureau, Business Trends and Outlook SurveyOctober 22, 2026www.census.gov/hfp/btos/downloads/Sector.xlsx
  10. 10Weshalb die meisten Betriebe Künstliche Intelligenz (noch) nicht nutzenInstitut für Arbeitsmarkt- und Berufsforschung (IAB), IAB-Forum (C. Kagerl), DOI 10.48720/IAB.FOO.20261020.01October 20, 2026iab-forum.de/weshalb-die-meisten-betriebe-kuenstliche-intelligenz-noch-nicht-nutzen/
  11. 11Sound Practices for Responsible Adoption of Artificial Intelligence (AI): Final reportFinancial Stability Board (FSB)October 15, 2026www.fsb.org/2026/10/sound-practices-for-responsible-adoption-of-artificial-intelligence-ai-final-report/
  12. 12Business insights and impact on the UK economy confidence intervals: BICS Wave 165 edition (wave165confidenceintervals.xlsx)Office for National Statistics (ONS)October 8, 2026www.ons.gov.uk/economy/economicoutputandproductivity/output/datasets/businessinsightsandimpactontheukeconomyconfidenceintervals
  13. 13Baromètre France Num: Résultats de l’enquête 2026 (Rapport final)Direction générale des Entreprises (DGE) / France Num; Erhebung durch CrédocSeptember 17, 2026www.francenum.gouv.fr/files/2026-09/Barom%C3%A8tre%20France%20Num%202026%20-%20Rapport%20final.pdf
  14. 14Künstliche Intelligenz in deutschen Betrieben: Jeder vierte Betrieb nutzt mittlerweile generative KI (IAB-Kurzbericht 8/2026)Institut für Arbeitsmarkt- und Berufsforschung (IAB), M. Friedrich und C. Kagerl, DOI 10.48720/IAB.KB.2608May 5, 2026doku.iab.de/kurzber/2026/kb2026-08.pdf
  15. 15AI is gaining ground in French firms: rapid diffusion but productivity gains are so far limited (Bulletin de la Banque de France 266/1)Banque de France (V. Genre, L. Parpais)September 3, 2026www.banque-france.fr/en/publications-and-statistics/publications/ai-gaining-ground-french-firms-rapid-diffusion-productivity-gains-are-so-far-limited
  16. 16Enterprises' use artificial intelligence (10+ employees) by enterprise size, industry (DB25 selected industries) and use (StatBank-Tabelle ITAV28)Statistics Denmark (Danmarks Statistik)September 15, 2026www.statbank.dk/ITAV28
  17. 172026 State of Nonprofit AI Adoption and Governance ReportNTEN und The Bridgespan GroupSeptember 10, 2026www.nten.org/publications/state-of-nonprofit-ai
  18. 18Artificial intelligence adoption in Chinese enterprises: Productivity effects, workforce implications, and policy challenges (Research brief)International Labour Organization (E. Ernst, Z. Zhao, H. Zhu, Y. Li, X. Wei, H. Zhang, Z. Chen), DOI 10.54394/00035342September 15, 2026www.ilo.org/publications/artificial-intelligence-adoption-chinese-enterprises-productivity-effects
  19. 19ICT usage in enterprises (statistics release 2026, Table 1)Statistics Norway (SSB)September 25, 2026www.ssb.no/en/teknologi-og-innovasjon/informasjons-og-kommunikasjonsteknologi-ikt/statistikk/bruk-av-ikt-i-naeringslivet
  20. 20A Simple Solution to Improving Human Supervision of Algorithms: Evidence from Smart Vending (arXiv 2607.00420, Version 2)M. Zhao, B. R. Han, X. Chen, T. Zhu (arXiv-Preprint, cs.HC)July 1, 2026arxiv.org/abs/2607.00420v2
  21. 21Bruken av KI i næringslivet akselerererStatistisk sentralbyrå (SSB), M. RybalkaSeptember 25, 2026www.ssb.no/teknologi-og-innovasjon/informasjons-og-kommunikasjonsteknologi-ikt/statistikk/bruk-av-ikt-i-naeringslivet/artikler/bruken-av-ki-i-naeringslivet-akselererer
Authors, reuse basis and rights records
  1. 1U.S. Census Bureau, Business Trends and Outlook Survey · Reuse basis: Work of U.S. Census Bureau employees, not subject to copyright in the United States (17 U.S.C. § 105; Census policy DS027); rights may exist abroad; our own paraphrase and link, no tables reproduced · Rights and license record
  2. 2Office for National Statistics (ONS) · Reuse basis: Open Government Licence v3.0; our own paraphrase and link, no figures or tables reproduced · Rights and license record
  3. 3Statistics Canada (B. Li, R. Fair, C. Johnston), Catalogue no. 11-621-M · Reuse basis: Statistics Canada Open Licence; our own paraphrase with attribution and link, no tables reproduced · Rights and license record
  4. 4Instituto Nacional de Estadística (INE), Spain · Reuse basis: CC BY 4.0 (INE general licence for statistical information); attribution 'Source: INE' with link, our own paraphrase, no tables reproduced · Rights and license record
  5. 5European Commission, Directorate-General for Communication (Eurobarometer) · Reuse basis: Reuse under Commission Decision 2011/833/EU, content under CC BY 4.0; our own analysis and paraphrase with attribution, no graphics; the Commission is not responsible for this analysis · Rights and license record
  6. 6European Commission, Joint Research Centre (M. Ignaszak, D. Nepelski, J. Stake, A.-A. André) · Reuse basis: CC BY 4.0 (reuse under Commission Decision 2011/833/EU, per the report's imprint); our own paraphrase with attribution, no figures; third-party cover image excluded · Rights and license record
  7. 7Statistik Austria (Bundesanstalt Statistik Österreich) · Reuse basis: Copyright STATISTIK AUSTRIA; according to the information sheet of the table volumes, reproduction and distribution for commercial and non-commercial purposes is permitted with correct rendering and the source credit "STATISTIK AUSTRIA", edits must be marked; our own paraphrase and link, no tables reproduced · Rights and license record
  8. 8U.S. Census Bureau, Business Trends and Outlook Survey · Reuse basis: Work of U.S. Census Bureau employees, not subject to copyright in the United States (17 U.S.C. § 105; Census policy DS027); rights may exist abroad; our own paraphrase and link, no tables reproduced · Rights and license record
  9. 9U.S. Census Bureau, Business Trends and Outlook Survey · Reuse basis: Work of U.S. Census Bureau employees, not subject to copyright in the United States (17 U.S.C. § 105; Census policy DS027); rights may exist abroad; our own paraphrase and link, no tables reproduced · Rights and license record
  10. 10Institut für Arbeitsmarkt- und Berufsforschung (IAB), IAB-Forum (C. Kagerl), DOI 10.48720/IAB.FOO.20261020.01 · Reuse basis: CC BY-SA 4.0 (per the article page); our own paraphrase with attribution and link, no figures reproduced · Rights and license record
  11. 11Financial Stability Board (FSB) · Reuse basis: Copyright 2026 Financial Stability Board; under its terms, non-commercial reproduction with attribution and without modification is permitted; here only our own paraphrase and link · Rights and license record
  12. 12Office for National Statistics (ONS) · Reuse basis: Open Government Licence v3.0; our own paraphrase and link, no tables reproduced · Rights and license record
  13. 13Direction générale des Entreprises (DGE) / France Num; Erhebung durch Crédoc · Reuse basis: Licence Ouverte / Etalab 2.0 for content produced by the DGE (per the francenum.gouv.fr legal notice); attribution with date, our own paraphrase, no charts reproduced · Rights and license record
  14. 14Institut für Arbeitsmarkt- und Berufsforschung (IAB), M. Friedrich und C. Kagerl, DOI 10.48720/IAB.KB.2608 · Reuse basis: CC BY-SA 4.0 (per the imprint of the brief); our own paraphrase with attribution and link, no figures reproduced · Rights and license record
  15. 15Banque de France (V. Genre, L. Parpais) · Reuse basis: Copyright Banque de France; under its legal notice, data whose sole source is the Banque de France may be reused commercially and non-commercially with source and date and without distortion; our own paraphrase and link, no charts reproduced · Rights and license record
  16. 16Statistics Denmark (Danmarks Statistik) · Reuse basis: Free reuse including commercial use with the source credit "Statistics Denmark"; according to Statistics Denmark this corresponds to CC BY 4.0, processing must be indicated; our own paraphrase and link, no tables reproduced · Rights and license record
  17. 17NTEN und The Bridgespan Group · Reuse basis: Copyright 2026 NTEN; no open licence stated; only our own factual paraphrase of individual results and a link, no figures reproduced · Rights and license record
  18. 18International Labour Organization (E. Ernst, Z. Zhao, H. Zhu, Y. Li, X. Wei, H. Zhang, Z. Chen), DOI 10.54394/00035342 · Reuse basis: CC BY 4.0 (per the imprint of the research brief, International Labour Organization 2026); our own paraphrase with attribution, no tables reproduced · Rights and license record
  19. 19Statistics Norway (SSB) · Reuse basis: CC BY 4.0 (per the SSB licence page); attribution to Statistics Norway with link, our own paraphrase, no tables reproduced · Rights and license record
  20. 20M. Zhao, B. R. Han, X. Chen, T. Zhu (arXiv-Preprint, cs.HC) · Reuse basis: CC BY-NC-SA 4.0 (per the arXiv page); only our own factual paraphrase and link, no text or figures reproduced · Rights and license record
  21. 21Statistisk sentralbyrå (SSB), M. Rybalka · Reuse basis: CC BY 4.0 (per the SSB licence page); attribution to Statistics Norway with link, our own paraphrase, no figures reproduced · Rights and license record

How to cite

qasb Research (2026). AI in organizations: more businesses use AI. Topic report 10/2026, data cutoff October 26, 2026. https://www.qasb.ai/en/reports/2026-10-organisationen

Appendix

For every statement: evidence type, change status, scope and limitations, and the observation period, source value, comparison and uncertainty where the source gives them. The texts are the approved entries of the edition, as is its summary.

Scope and limitations of all 23 statements (A1–A23)

Edition summary

Official surveys from the United States, the United Kingdom, Austria, Denmark and Norway show further rising shares of businesses using AI in 2026 (in Canada planned use is rising; for Spain no comparable previous-year value exists), with a wide gap between small and large firms and concentrations in information, finance and in marketing and sales; many users have so far spent little or nothing on AI, and about a third of UK businesses with ten or more employees that use AI or are unsure have trained no staff. Establishments without AI mainly cite missing use cases; missing knowledge, a lack of skilled staff and legal uncertainty are each named by about a third. Effects on productivity and employment rest mostly on self-reports, only one older randomised study covers a narrow use case, and country figures are not directly comparable because questions and size classes differ.

  1. A1Back to chapter 2.0

    Evidence: observed · baseline · Source 1

    Scope
    Business Trends and Outlook Survey: a biweekly business survey with weighted national estimates; question on AI use in any business function in the last two weeks. In the same file, 28.6% expected to use AI in the next six months.
    Limitations
    Self-reported use; says nothing about intensity, benefit or employment effects. The 22 October 2026 file relabels AI as 'Super Intelligence (SI), formerly AI', citing Executive Order 14434, including for earlier periods; it does not say whether respondents saw changed question wording. Not comparable with other countries' surveys because questions and definitions differ.
    Observation period
    Reference period 21 September–4 October 2026 (collection 5–18 October 2026, published 22 October 2026); comparison value from collection 17–30 November 2025
    Source value
    24.6% of businesses (weighted estimate, AI use in the last two weeks)
    Baseline
    17.3% in the first period of this question series (collection 17–30 November 2025)
    Comparison
    Same question in the biweekly time series; previous period 23.8% (reference period 7–20 September 2026); descriptive, no impact comparison
    Uncertainty
    Standard error 0.27 percentage points (Census table 'Response Standard Errors')

  2. A2Back to chapter 2.0

    Evidence: observed · baseline · Sources 2, 12

    Scope
    Business Insights and Conditions Survey, wave 165 (21 September–4 October 2026), 9,933 responses from 38,592 sampled businesses (25.7%), weighted estimates; private sector excluding, among others, agriculture, energy generation, finance and insurance, public administration, and public education and health. Most common technologies (10+ employees): text generation with large language models (21%) and visual content creation (19%).
    Limitations
    Voluntary survey, labelled by the ONS as official statistics in development. The share measures diffusion, not intensity, quality or productivity effects; among businesses with fewer than ten employees it was 29%.
    Observation period
    Wave 165: 21 September–4 October 2026 (figures for September 2026); comparison September 2023
    Source value
    30% of covered businesses (weighted, at least one AI technology)
    Baseline
    9% in September 2023 (when the question was introduced)
    Comparison
    Same BICS question over time; descriptive
    Uncertainty
    No uncertainty in the bulletin; according to the ONS dataset with 95% confidence intervals (wave 165), 64.1% (61.4–66.9%) used no AI technology and 6.2% (4.8–7.5%) were not sure; among businesses with 10+ employees, 50.8% (48.8–52.8%) used no AI and 11.1% (9.9–12.4%) were not sure

  3. A3Back to chapter 2.0

    Evidence: observed · baseline · Source 16

    Scope
    StatBank table ITAV28 (updated 15 September 2026) from the EU-harmonised ICT survey; private non-agricultural activities excluding mining and excluding financial and insurance activities. By industry: IT and media 89%, manufacturing and utilities 56%, trade, transport, accommodation and restaurants 55%, construction 43%. Most common uses: text analysis 45%, production of written text or speech 44%, generation of pictures, video or audio 32%, workflow automation 25%.
    Limitations
    Use of at least one AI technology without information on intensity or effects; the table gives no uncertainty. Industry coverage differs from other countries (among other things, no financial sector).
    Observation period
    Survey 2026 (time series 2023–2026); table updated 15 September 2026
    Source value
    59% of enterprises with at least ten employees (AI use)
    Baseline
    15% in 2023; 42% in 2025
    Comparison
    10–49 employees 55%; 50–99 employees 69%; 100–249 employees 79%; 250+ employees 88% (2026)
    Uncertainty
    No uncertainty given in the table

  4. A4Back to chapter 2.0

    Evidence: observed · baseline · Source 7

    Scope
    ICT usage in enterprises survey 2026 under EU Regulation 2019/2152, stratified voluntary sample, about 6,200 participating enterprises, fieldwork February to July 2026. By industry: manufacturing 49.0%, trade 47.4%, construction 35.6%, telecommunications and IT 92.2%. The financial and insurance sector, included for the first time in 2026 (shortened questionnaire, separate table volume), reached 79.0%.
    Limitations
    The share measures use of at least one AI technology, not intensity, benefit or employment effects. Bracketed values have a sampling error above 5% according to the source; no confidence intervals are published. The value including the financial sector (48.7%) is only partly comparable with the previous year because of the shortened questionnaire.
    Observation period
    Fieldwork February to July 2026 (reference: time of survey); comparison 2025; tables created 20 October 2026
    Source value
    48.1% of enterprises with at least ten employees (excluding the financial and insurance sector)
    Baseline
    30% in 2025 (per the statistics page)
    Comparison
    10–49 employees 44.0%; 50–249 employees 65.4%; 250+ employees 84.7% (2026)
    Uncertainty
    No confidence intervals in the source; bracketed values have a sampling error above 5%

  5. A5Back to chapter 2.0

    Evidence: observed · baseline · Sources 19, 21

    Scope
    ICT usage in enterprises survey 2026 (Table 1 and analysis article of 25 September 2026), enterprises with ten or more employees in industries C to K, M to O and 95.1 (SIC2025), sample stratified by industry and size. According to SSB, the annual increase doubled to almost 20 percentage points compared with about 10 percentage points per year in 2023 to 2025; information and communication about nine in ten, transport and storage about three in ten.
    Limitations
    Size and industry figures in the article are rounded ('x in 10'), without uncertainty; the financial sector has not been covered since 2014. According to SSB the SIC2007 time series ended in 2025 and a new SIC2025 series starts in 2026, so the comparison with the previous year spans a classification change. Use without information on intensity or effects.
    Observation period
    Survey 2026; comparison 2025 and increases 2023–2025; published 25 September 2026
    Source value
    48% of enterprises with at least ten employees (at least one AI technology)
    Baseline
    About three in ten enterprises in 2025 (per the SSB article)
    Comparison
    Large enterprises (100+ employees) about eight in ten, small ones (10–19) about four in ten (2026)
    Uncertainty
    No uncertainty given in the source

  6. A6Back to chapter 2.0

    Evidence: observed · baseline · Source 4

    Scope
    Survey on ICT and e-commerce use in enterprises (following Eurostat guidance), stratified sample of about 15,000 enterprises with ten or more employees and about 10,000 smaller ones; AI questions refer to the first quarter of 2026. By sector (10+ employees): services 38.2%, industry 29.1%, construction 22.6%; for enterprises with fewer than ten employees INE reports 21.4%.
    Limitations
    Because of the switch to CNAE-2025 and a wider scope, INE says the data are not strictly comparable with earlier years. A usage share says nothing about intensity, benefit or employment effects.
    Observation period
    First quarter of 2026 (AI use); final data published 22 October 2026
    Source value
    33.6% of enterprises with at least ten employees
    Baseline
    No prior-year value in the release; INE says not strictly comparable with earlier years
    Comparison
    Descriptive level without impact comparison
    Uncertainty
    No uncertainty given in the source

  7. A7Back to chapter 2.0

    Evidence: observed · baseline · Source 3

    Scope
    Canadian Survey on Business Conditions, collected 2 July–6 August 2026; sample 21,161, results from 9,601 responses, calibrated weights. Most common planned applications among planning businesses: data analytics (41.7%), large language models (39.3%), virtual agents or chatbots (31.8%).
    Limitations
    Planned rather than measured use; the twelve-month window depends on the response date; no margins of error in the article. Statistics Canada reports that by the second quarter of 2026, 19.2% of businesses had used AI in the previous twelve months (different question and survey wave).
    Observation period
    Collection 2 July–6 August 2026 (Q3 2026); planning horizon next twelve months; comparison Q3 2025 and Q3 2024
    Source value
    25.2% of businesses (planned AI use in the next twelve months)
    Baseline
    14.5% in Q3 2025; 10.6% in Q3 2024
    Comparison
    Same question in the Q3 surveys 2024–2026; descriptive
    Uncertainty
    No uncertainty given in the source

  8. A8Back to chapter 3.0

    Evidence: observed · baseline · Source 8

    Scope
    Business Trends and Outlook Survey, table by employment size class, same question as the national series (AI use in any business function), reference period 21 September–4 October 2026. Intermediate classes: 10–19 employees 24.6%, 20–49 26.0%, 50–99 29.0%, 100–249 35.3%. In the first period of the question series (collection 17–30 November 2025) the values were 17.0% (1–4) and 30.5% (250+).
    Limitations
    Firm-level self-reports without information on intensity or effects. The file labels AI as 'Super Intelligence (SI)' (Executive Order 14434). Mid-sized classes have larger standard errors (up to 1.97 percentage points for 100 to 249 employees).
    Observation period
    Reference period 21 September–4 October 2026 (collection 5–18 October 2026, published 22 October 2026); comparison collection 17–30 November 2025
    Source value
    43.6% of businesses with 250 or more employees (AI use in the last two weeks)
    Baseline
    24.3% for 1 to 4 employees (same period)
    Comparison
    Same question across size classes; 250+ employees 30.5% in the 17–30 November 2025 collection
    Uncertainty
    Standard error 1.51 percentage points (250+) and 0.32 percentage points (1–4) per the Census table 'Response Standard Errors'

  9. A9Back to chapter 3.0

    Evidence: observed · baseline · Source 9

    Scope
    Business Trends and Outlook Survey, table by NAICS sector (52, 62, 31–33, 44–45), reference period 21 September–4 October 2026. Highest values: information 46.4%, professional, scientific and technical services 43.8%; lowest: agriculture and forestry 6.0%, accommodation and food services 10.7%. In the 17–30 November 2025 collection, finance and insurance stood at 29.9% and manufacturing at 12.7%.
    Limitations
    Self-reports by businesses; sector shares say nothing about depth of use or effects. Health care and social assistance combines hospitals, practices and social services; there is no separate figure for hospitals. Standard errors of the values cited range from 0.73 to 1.53 percentage points.
    Observation period
    Reference period 21 September–4 October 2026 (published 22 October 2026); comparison collection 17–30 November 2025
    Source value
    40.5% of businesses in finance and insurance (AI use in the last two weeks)
    Baseline
    29.9% in the 17–30 November 2025 collection
    Comparison
    Health care and social assistance 27.6%; manufacturing 22.9%; retail trade 18.5% (same period)
    Uncertainty
    Standard error 1.53 percentage points (finance), 0.88 (health care and social assistance), 1.35 (manufacturing), 0.73 (retail)

  10. A10Back to chapter 3.0

    Evidence: observed · baseline · Source 7

    Scope
    Base: an estimated 19,411 enterprises with at least ten employees using AI (excluding the financial sector), multiple answers, fieldwork February to July 2026. Large enterprises with 250+ employees used AI much more often for IT security (56.5% vs 15.3% for 10 to 49 employees); in trade, marketing or sales reached 80.5%.
    Limitations
    Data on areas of use, not on scale, cost or results; says nothing about whether AI supports or takes over tasks in these areas.
    Observation period
    Fieldwork February to July 2026; tables created 20 October 2026
    Source value
    67.5% of enterprises using AI (area of use: marketing or sales)
    Baseline
    All Austrian enterprises with 10+ employees using AI in 2026 (excluding the financial sector)
    Comparison
    Management/organisation 51.4%; production/service processes 40.6%; logistics 9.2%
    Uncertainty
    No uncertainty given in the source

  11. A11Back to chapter 4.0

    Evidence: observed · baseline · Source 10

    Scope
    IAB Establishment Panel, second half of 2025; base: establishments neither using nor planning AI (about two thirds of all establishments). Almost 90% of micro-establishments named the lack of use cases, but fewer than two thirds of establishments with 200 or more employees; only 7% had tried AI without their expectations being met.
    Limitations
    Self-reports with multiple answers. According to the article the associations persist in multivariate regressions, but coefficients and sample sizes are not reported. Reasons are asked only of establishments without use or plans.
    Observation period
    Survey in the second half of 2025 (IAB Establishment Panel); published 20 October 2026
    Source value
    36% of establishments neither using nor planning AI (reason: missing knowledge)
    Baseline
    No use case: almost 90%
    Comparison
    Skilled staff 34%; incompatibility 33%; data protection and security 32%; legal uncertainty 32%; cost 30%
    Uncertainty
    No uncertainty given in the article

  12. A12Back to chapter 4.0

    Evidence: observed · baseline · Source 5

    Scope
    12,826 surveyed companies in the EU-27 with 1–749 employees (weighted; about 90% micro-enterprises with 1–9 employees), fieldwork 20 March–13 April 2026. 42% used AI or automated systems at the workplace (12% extensively, 30% in a limited or experimental way), and a further 21% planned to.
    Limitations
    Questions also cover automated systems without AI (algorithmic management) and refer to workforce-related uses. Self-reports by company heads; the high share of micro-enterprises shapes the EU figures. No measurement of actual costs or effects.
    Observation period
    Fieldwork 20 March–13 April 2026; published 14 October 2026
    Source value
    50% of surveyed companies (very or rather important difficulty: high investment costs)
    Baseline
    All respondents in the EU-27, weighted base 12,826
    Comparison
    Cross-section without previous wave; comparison items 49% (legal uncertainty) and 48% (lack of expertise)
    Uncertainty
    No uncertainty given in the source

  13. A13Back to chapter 4.0

    Evidence: observed · baseline · Source 6

    Scope
    EU firms with at least ten employees; ten strategic sectors of the Apply AI Strategy mapped via NACE weights; Eurostat surveys 2021 and 2023–2025 (reasons for non-use from 2023). Overall, the share of firms using at least one AI technology rose from 8% (2021) to 20% (2025), according to the JRC.
    Limitations
    Secondary analysis of official data; changes in barriers are relative shifts among firms that considered AI, not absolute rates; the sector mapping via NACE weights has coverage gaps. No causal explanation.
    Observation period
    Eurostat ICT surveys 2021, 2023, 2024 and 2025 (barriers 2023–2025); published 14 October 2026
    Source value
    70% of firms that considered but did not adopt AI (rounded; barrier: lack of expertise)
    Baseline
    Reasons for non-use in 2023 as the starting point of the change
    Comparison
    Relative change of barriers 2023–2025 by strategic sector
    Uncertainty
    No uncertainty given in the source; the report rounds the value to 'roughly 70%'

  14. A14Back to chapter 5.0

    Evidence: observed · baseline · Source 12

    Scope
    Business Insights and Conditions Survey, wave 165 (fieldwork 21 September–4 October 2026), dataset with 95% confidence intervals of 8 October 2026; weighted shares; base: businesses not permanently stopped trading that use AI or are unsure whether they do. Further bands: £50,000 to £99,999 2.6%, £100,000 to £499,999 2.0%. Among businesses with 250+ employees, 57.4% were not sure.
    Limitations
    Self-reports in value bands without exact amounts; the high 'not sure' share limits interpretation. Voluntary survey labelled by the ONS as official statistics in development; finance and insurance and other sectors are not covered.
    Observation period
    Wave 165: fieldwork 21 September–4 October 2026, investment over the last twelve months; comparison wave 159 (15–28 June 2026)
    Source value
    35% of businesses with 10+ employees using AI or unsure (investment below £10,000 in twelve months)
    Baseline
    23.4% made no investment in AI
    Comparison
    Wave 159 (June 2026): 32.2% below £10,000; descriptive
    Uncertainty
    95% confidence interval 32.3–37.6% (ONS dataset)

  15. A15Back to chapter 5.0

    Evidence: observed · baseline · Source 13

    Scope
    Online and telephone survey of 9,655 businesses (23 March–18 April 2026) by Crédoc for the Directorate General for Enterprise (DGE), weighted by size, sector and region. Most common uses: text, voice or image generation 34%, chatbots and assistants 24%, document analysis 14%, data analysis or forecasting 11%, task automation 9%. By size: 1 to 4 employees 36%, 50 to 249 employees 62%. Further spending by AI users: training 12%, own developments 7%.
    Limitations
    Self-reports; the spending question (new in 2026) records no amounts. The usage share includes occasional use of free tools and, because micro-enterprises are included, is not comparable with the EU ICT survey (ten or more employees).
    Observation period
    Fieldwork 23 March–18 April 2026; comparison 2024; published 17 September 2026 (report version of 2 October 2026)
    Source value
    40% of businesses with 0 to 249 employees (AI use)
    Baseline
    13% in 2024
    Comparison
    Micro-enterprises (TPE) 38%; PME 53%; AI users with spending on subscriptions or licences 43%
    Uncertainty
    Margin of error below 1% according to the source (95% level, total sample)

  16. A16Back to chapter 5.0

    Evidence: observed · baseline · Source 12

    Scope
    Business Insights and Conditions Survey, wave 165 (fieldwork 21 September–4 October 2026), confidence interval dataset of 8 October 2026; weighted shares; base as for the investment question. Across all size bands 47.7% had trained no one; among businesses with 250+ employees it was 12.5%, with 45.9% 'not sure'.
    Limitations
    Self-reports; 'training' is a coarse measure without information on content, duration or quality. Voluntary survey (official statistics in development).
    Observation period
    Wave 165: fieldwork 21 September–4 October 2026
    Source value
    33.9% of businesses with 10+ employees using AI or unsure (no staff trained)
    Baseline
    All size bands: 47.7% without training
    Comparison
    250+ employees: 12.5% without training; descriptive
    Uncertainty
    95% confidence interval 31.2–36.6% (ONS dataset)

  17. A17Back to chapter 5.0

    Evidence: observed · baseline · Source 5

    Scope
    Using companies in the EU-27 (unweighted 5,726, weighted 5,357) with 1–749 employees, fieldwork 20 March–13 April 2026; a further 7% reported system decisions with human review only after a complaint.
    Limitations
    Self-reports without a breakdown by decision type (for example scheduling or performance assessment); also covers systems without AI. The answers show mostly partial automation with a final human decision, but nothing about decision quality.
    Observation period
    Fieldwork 20 March–13 April 2026; published 14 October 2026
    Source value
    2% of using companies (fully automated decisions)
    Baseline
    52%: systems suggest, humans always decide
    Comparison
    Cross-section without previous wave
    Uncertainty
    No uncertainty given in the source

  18. A18Back to chapter 5.0

    Evidence: observed · baseline · Source 14

    Scope
    IAB Establishment Panel 2025 (about 15,000 establishments of all sizes with at least one employee subject to social security contributions each year, fieldwork 1 July–2 November 2025, extrapolated); just under 25% of all establishments used generative AI. 90% of user establishments used publicly available models, 16% purchased models trained on their own data, 6% their own models.
    Limitations
    Published on 5 May 2026, before the reporting period. Spending without amounts; rules and training without information on content or bindingness; use by employees without the establishment's knowledge is not captured.
    Observation period
    Fieldwork 1 July–2 November 2025; spending in the last financial year (usually 2024); published 5 May 2026
    Source value
    45% of establishments using generative AI (spending on generative AI in the last financial year)
    Baseline
    Just under 25% of all establishments used generative AI in 2025
    Comparison
    Internal rules 21% (a further 22% planned); training 27% (a further 21% planned)
    Uncertainty
    No confidence intervals given in the brief for these shares

  19. A19Back to chapter 5.0

    Evidence: observed · baseline · Source 17

    Scope
    Survey of staff and executives of non-profit organisations in the United States and worldwide in summer 2026 with 917 responses (per the report page); budget and governance answers come from executives only. A further 23% said a dedicated AI budget was in development; a further 45% and 44% were working on a risk plan and on training for safe use, respectively.
    Limitations
    Non-representative convenience sample with self-selected participants and self-reports; the publishers are a professional network and a consultancy for non-profits. The full report is available only after a free NTEN sign-in; the public report page and the charts offered there for sharing were checked. Results cannot be transferred to public enterprises or other countries.
    Observation period
    Survey in summer 2026 (per the report page); published 10 September 2026
    Source value
    57% of surveyed executives (no dedicated AI budget)
    Baseline
    17% with a dedicated AI budget
    Comparison
    Named AI oversight 42%; risk plan 22%; training on safe use 22%
    Uncertainty
    No uncertainty given; non-representative sample

  20. A20Back to chapter 5.0

    Evidence: observed · baseline · Source 11

    Scope
    Final report after public consultation, addressed to all types of financial institutions, with illustrative case studies; applied proportionately, with more robust practices for large, complex or highly connected institutions and AI in critical functions.
    Limitations
    Not a binding international standard and no measurement of actual AI use or governance in institutions; according to the FSB the case studies do not describe expected outcomes. How national supervisors take up the practices is open.
    Observation period
    Published 15 October 2026

  21. A21Back to chapter 6.0

    Evidence: observed · baseline · Source 15

    Scope
    Supplementary questions in the Monthly Business Survey, 25 February–3 April 2026; manufacturing, construction and non-financial market services excluding wholesale and retail trade, weighted by size and sector. Generative AI: 64% of firms with 20 to 49 employees, 83% of firms with 1,000+; non-generative AI about one third, of which 41% reported productivity gains. Net balance of AI-related investment increases: 15% (generative AI) and 22% (non-generative).
    Limitations
    Self-reports and net balances, no measured productivity data; firms with fewer than 20 employees are excluded. Expectations for the next three years are much more optimistic than the effects observed so far.
    Observation period
    Fieldwork 25 February–3 April 2026; published 3 September 2026
    Source value
    31% of firms using generative AI (observed productivity gains)
    Baseline
    67% of firms with 20+ employees use generative AI
    Comparison
    Non-generative AI: 41% with productivity gains; workforce effect net 2% decline (generative AI)
    Uncertainty
    No uncertainty given in the source

  22. A22Back to chapter 6.0

    Evidence: observed · baseline · Source 18

    Scope
    Qualitative interviews by Renmin University with 21 firms in manufacturing, finance, business services, construction, education, media and travel, complemented by a survey of 1,591 professionals; published 15 September 2026. Reported individual figures range from 30 to 100% efficiency gains in core workflows to a 57% shorter recruitment cycle.
    Limitations
    Small, non-representative sample; productivity figures come from managers, are not independently verified and are not comparable with each other. According to the brief the interviews took place in the first half of 2026; no period is stated for the survey of professionals.
    Observation period
    Interviews in the first half of 2026 (period of the professionals survey not stated); published 15 September 2026

  23. A23Back to chapter 6.0

    Evidence: observed · baseline · Source 20

    Scope
    Random assignment to no, free or constrained override of autonomous AI replenishment; according to the full text the partner was the vending operator Fengyi Technology (China). arXiv preprint, first version 1 July 2026, revised version 17 October 2026. The constrained override also worked for experienced workers and top-selling items without significant sales losses.
    Limitations
    Non-peer-reviewed preprint on a narrow use case (inventory planning for vending machines) at a single firm in China. First version predates the reporting period; evidence on designing partial automation with human control, not on productivity in general.
    Observation period
    Field experiment (period not stated in the abstract); first version 1 July 2026, version 2 of 17 October 2026
    Source value
    1.28 log points lower inventory (constrained override vs no override)
    Baseline
    No override possible
    Comparison
    Free override: inventory minus 1.95, sales minus 1.19 log points
    Uncertainty
    Sales decline under constrained override not statistically significant; intervals only in the full text

And in your case? Check it with sources.

Start an analysisFor organizations
All reports