What can 816 Slovak organisations tell us about AI readiness?
A new peer-reviewed data resource documents AI use, plans, sentiment and barriers across 816 deduplicated Slovak organisations. Its value is a transparent baseline for secondary research—not a representative national adoption rate or proof that AI improves work.
Editorial responsibility: The Impact of AI Editorial Desk · Report a factual concern
At a glance
- 1Researchers fielded an online computer-assisted web interview in Slovakia from 18 August to 3 September 2025 and retained 816 unique organisation records after screening and deduplication.
- 2The dataset covers organisational and respondent characteristics, AI maturity and sentiment, current and planned functional use, and perceived barriers, with raw and cleaned tables plus survey documentation.
- 3Because this is a reusable dataset rather than an official probability estimate or causal evaluation, readers should not convert its responses into a national adoption rate or evidence that AI caused performance changes.
Research topic
How an open Slovak enterprise survey can support analysis of organisational AI maturity, use, plans and barriers
The direct answer: it offers a detailed baseline, not a national score
The 816 Slovak organisations can tell researchers how a defined set of established enterprises described their AI use, plans, attitudes and barriers during a short 2025 field period. The newly published Data Descriptor makes that material reusable by releasing documented raw and cleaned survey files. It is valuable because workplace-AI debate often runs ahead of comparable organisation-level evidence.
It cannot, by itself, tell us the share of every Slovak business using AI. The final file is the result of eligibility screening and deduplication, not a complete register of enterprises or a causal experiment. The authors' contribution is a research dataset whose structure, questions and cleaning decisions can be inspected. Any prevalence claim needs an appropriate sampling design, weights and uncertainty analysis rather than dividing one response count by another.
The distinction matters for workers and policymakers. A well-documented survey can reveal where organisations report capability gaps or planned uses and can support comparisons within the sample. It does not show that an announced tool is operational, that employees use it safely, that productivity changed or that benefits and risks are evenly distributed across jobs.[1]
How the 816-record dataset was assembled
The team administered a computer-assisted web interview, or CAWI, in Slovakia from 18 August to 3 September 2025. Respondents first had to confirm that the organisation operated in Slovakia, had at least 10 employees, annual turnover of at least €200,000 and at least five years of operation. State-owned enterprises and state organisations were excluded, while mixed-ownership organisations could remain.
The screening produced 1,086 complete responses. The researchers then deduplicated organisations, leaving 816 unique records for the analytical dataset. That two-stage denominator should stay visible: 1,086 is the number of completed eligible questionnaires before organisation-level deduplication; 816 is the number of retained unique organisations. Treating both as independent enterprises would overstate the organisational sample.
The eligibility rules deliberately focus on established organisations of some minimum scale. Very small firms, new ventures, low-turnover businesses and state bodies are outside the frame. That may improve the relevance of questions about structured strategy and functional deployment, but it narrows what the results can say about the Slovak economy as a whole.[1]
What is inside the resource
The questionnaire spans organisation and respondent characteristics, perceived AI maturity, sentiment toward AI, current adoption, intended adoption, functional uses and perceived barriers. This breadth allows analysts to study relationships between an organisation's context and the way its representative describes readiness. It also supports separating current use from plans, a basic distinction often blurred in corporate surveys.
The published package includes raw and cleaned data tables, questionnaire materials, a skip-logic map and a variable dictionary. Those components make the research more useful than a press-release percentage. A secondary analyst can see which respondents received which questions, how variables are labelled and how the final file differs from the initial responses.
Documentation does not make every answer objectively true. One representative may have incomplete knowledge of activity across departments, and terms such as adoption, maturity or readiness can be interpreted differently. The dataset records reported organisational conditions at the time of fieldwork. It is not an inventory validated against software logs, procurement records or employee observations.[1]
Why deduplication and skip logic matter
Organisation surveys can receive more than one response from the same employer. Without deduplication, a large or highly engaged enterprise could appear several times and exert too much influence. Reducing 1,086 complete responses to 816 unique organisations addresses that problem, although users still need to inspect how duplicates were identified and which record was retained before reproducing an analysis.
Skip logic is equally important. A respondent who reports no current AI use should not necessarily answer detailed questions about operational deployments, while someone reporting active use may follow a longer path. If an analyst interprets structurally missing answers as ‘no’, estimated barriers or functional use could be badly distorted. Publishing the logic map helps prevent that mistake.
Raw and cleaned versions also make quality decisions auditable. Researchers can test whether conclusions change when they use alternative treatments of missing values, ambiguous answers or organisational duplicates. This is one reason a Data Descriptor is consequential even when it does not announce a single headline effect: it expands the evidence that others can check, combine and challenge.[1]
What employers, workers and policymakers can use it for
Employers can use the question set as a diagnostic prompt: do they know which functions use AI, whether activity is experimental or routine, what staff believe the barriers are and where planned deployment lacks skills or governance? They should not use the file as a league table. An organisation's self-described maturity can reflect confidence and vocabulary as much as tested capability.
For workers, the useful follow-up is whether adoption changes tasks, workload, discretion, pay, monitoring and training. The released themes can help researchers identify organisations or sectors for deeper study, but a management respondent cannot stand in for employee experience. Linking organisation-level answers to confidential worker surveys or qualitative interviews would reveal whether strategic intent matches practice.
Policymakers can examine how barriers vary within this eligible sample and design hypotheses about support, infrastructure or skills. Before allocating resources, they need representative estimates, transparent weighting and outcome evidence. A reported intention to adopt AI may never become a deployment; a deployment may remain a small pilot; and an active system may create costs as well as benefits.[1]
The main limits: selection, self-report and time
The online mode is efficient but can select for people and organisations willing to complete a digital survey about AI. The published resource should therefore be analysed with its recruitment and eligibility conditions in view. Unless a secondary analysis demonstrates population coverage and suitable weights, percentages describe the responding dataset rather than every enterprise in Slovakia.
Self-report introduces a second uncertainty. Respondents may overstate readiness, define AI broadly or narrowly, or answer according to planned strategy rather than verified use. Social desirability can work in both directions: some may present their organisation as innovative, while others may under-report informal or unapproved tools. Independent operational measures would be needed to validate deployment.
The fieldwork took place over 17 days in August and September 2025, while the Data Descriptor was published on 7 October 2026. The original observation date should therefore remain attached to any analysis. The resource is newly available, but its answers are not a live October 2026 pulse. Fast-changing model availability and regulation may have altered organisational plans since collection.[1]
Funding, geography and responsible comparison
The project was funded through the European Union's NextGenerationEU recovery instrument in Slovakia under project 09I05-03-V02-00003/2025/VA, identified as AI-impactSK. The authors report no competing interests. Public funding and disclosure support transparency, while the data and methods still need to be judged independently.
Slovakia offers an important Central European setting that is often missing from AI-adoption evidence dominated by the United States, Western Europe or global convenience samples. Its industrial structure, firm sizes, labour market and public-support programmes shape the responses. Cross-country comparison will require harmonised questions and equivalent sampling, not juxtaposing unrelated headline percentages.
The exclusion of state organisations also means the dataset should not be used to infer public-sector readiness. Mixed-ownership organisations remained eligible, so ownership variables need careful handling. Analysts should state which subset they use and retain the enterprise eligibility thresholds whenever they communicate findings.[1]
What evidence would change the assessment
The next advance would be a probability-based or well-calibrated longitudinal survey that follows the same organisations and reports response rates, weights and sampling uncertainty. Repeated waves could separate temporary enthusiasm from durable adoption and show whether planned functions actually move into routine use.
Validation should combine organisational responses with concrete measures: approved-system inventories, expenditure, task-level use, training hours, incident reports and employee accounts. Outcome studies could then compare matched organisations or phased deployments, measuring productivity, quality, job design, wages and worker wellbeing rather than assuming readiness produces benefits.
For now, the responsible reading is modest but useful. The 816-record dataset creates a transparent foundation for questions about AI readiness in established Slovak organisations. Its importance lies in what researchers can test next—not in turning a documented convenience of observation into a definitive national adoption figure.[1]
What this means for people
- Workers gain visibility only indirectly: the dataset maps organisational claims and barriers, but does not substitute for employee-reported experience.
- Employers receive a reusable question set for examining strategy, functions and barriers without gaining a benchmarking league table.
- Policymakers obtain a transparent research baseline but still need representative, longitudinal and outcome evidence before directing support.
Global context
The resource adds organisation-level evidence from Slovakia to a literature often centred on larger Western markets. It can support carefully harmonised regional research, but comparisons must preserve its August–September 2025 field period, private and mixed-ownership eligibility rules, minimum size and age thresholds, and self-reported design. Its contribution is an open basis for analysis rather than a country ranking.
What the evidence does not yet show
- The data are self-reported by organisational representatives and were not validated against system logs, procurement records or employee observations.
- Eligibility excluded organisations with fewer than 10 employees, turnover below €200,000, fewer than five years of operation, and state-owned or state organisations.
- The final 816 records are unique organisations after deduplication; they should not automatically be treated as a representative probability sample of Slovak enterprises.
- Fieldwork ran from 18 August to 3 September 2025, so publication on 7 October 2026 does not make the observations contemporaneous with publication.
- The resource is descriptive and cross-sectional; it cannot show that AI adoption caused productivity, workforce or business outcomes.
What to watch next
- Secondary analyses that document recruitment, weights, uncertainty and missing-data treatment before estimating prevalence.
- Longitudinal waves showing whether reported plans became sustained organisational use.
- Linkage to employee surveys, task evidence, training records and independently verified deployments.
- Comparable surveys in Central and Eastern Europe using harmonised questions and sampling frames.
- Outcome studies measuring job quality, productivity, safety and distributional effects rather than readiness alone.
Living evidence record
Impact record IAI-1Q2IW6V
Evidence stage
Studied
Confidence
Supported
Reporting basis
Source analysis
Independent or research support
Present
Record status
Monitoring
Last checked
7 October 2026
Source trail
1 direct source across 1 source type.
People impact
Documented in this record.
Uncertainty
Limits and next checks are explicit.
Stages describe the evidence available—not whether a technology is good or bad. See the public method.
Single-source reporting disclosure
This record analyses one direct source. It can establish what Scientific Data published or reported, but it is not independent corroboration of every performance claim or predicted outcome. The confidence label will change only when broader evidence is added.
Evidence trail
Sources used for this report
Links checked 7 October 2026
This report is labelled source analysis. We summarise and analyse source material in our own words; company statements remain attributed claims until independently supported. Translated summaries preserve the meaning of the original source and link back to it. Read our editorial standards.
Continue the story
Related reporting
Work & Skills
Who is workplace AI leaving behind? PwC finds an access gap—but not its cause
PwC surveyed 49,364 workers across 48 countries and regions. AI use rose, while reported access to learning fell; the cross-sectional responses show an association, not proof that AI caused the divide.
5 min · 2 sources
Work & Skills
Who is gaining from AI at work?
A Gallup survey of 15,482 US employees finds that regular AI users report more speed, creativity and work quality, but frequent use is concentrated among graduates, managers and workers who already have better jobs. The results are self-reported associations, not a productivity trial.
8 min · 2 sources
Work & Skills
If robots can do much physical work, why are so few tasks economical to automate?
Anthropic researchers rate 7,594 physical job tasks and estimate that robots can perform 74% in some setting, yet are currently cheaper than people for only 0.3% of all work. The gap between technical exposure and economic adoption is the result that matters for workers.
6 min · 1 source
The Impact Brief
Keep the evidence trail, not the noise.
Get the most consequential AI developments with direct sources and clear limits.
Reader commentary
Add evidence, experience or a question
No account is required. Reader notes are published after a brief civility, relevance and safety check; disagreement is welcome.
Explore commentary across the portal →Published reader notes
0No published reader notes yet. You can start the evidence-led discussion above.
Prefer a private correction or response? Contact the newsroom.