AI-ImpactSK: Semi-structured interviews on AI adoption and implementation in Slovak enterprises (two waves, bilingual)

A bilingual (Slovak/English) qualitative dataset of 58 anonymised semi-structured interviews on the
adoption and implementation of artificial intelligence in Slovak enterprises and public organisations,
collected in two waves during 2025-2026 within the AI-ImpactSK project at the Faculty of Business
Management, Bratislava University of Economics and Business.

This record contains the cleaned transcripts of all 58 interviews in Slovak and in English
translation (116 files), anonymised participant metadata, a structured thematic codebook of 8 categories
and 24 themes with operational definitions and inclusion and exclusion criteria, segment-level thematic
coding of 629 segments, the interview protocol in both languages, a fieldwork report documenting
recruitment and response rates, the derived interview-level dataset with a full data dictionary,
machine-readable construct and index definitions, the inter-coder reliability table, and the complete
analysis code with pinned dependencies. Following code/RUN.md regenerates every statistic and every
figure reported in the accompanying article from the deposited files.

These are the supplementary materials for the article "Adoption ahead of governance: quantifying the
formalization gap in artificial intelligence use across Slovak enterprises".

Access. Everything in this record is openly available under CC BY 4.0; the analysis code in code/ is
licensed MIT. The verbatim (near-literal) transcripts are deposited separately, under restricted access,
at https://doi.org/10.5281/zenodo.22747631, and are available on request. They retain contextual detail,
such as product descriptions, client and partner relationships and named platforms and events, that
could permit re-identification of a participating organisation in a small national economy even after
all direct identifiers have been removed. The cleaned transcripts published here carry the same
substance.

Audio recordings are not shared: voice is an identifier, and the recorded consent covered transcription,
analysis and publication of anonymised outputs only. The anonymisation key linking placeholders to
original names is held by the authors and is not deposited.