Bibliometric Corpus and Cluster Data: Institutionally Conditioned Absorptive Capacity for AI-Enabled Strategy

Bibliometric Corpus and Cluster Data — "Institutionally Conditioned Absorptive Capacity for AI-Enabled Strategy" Author: Alfredo Agustín Merlet Echavarría (ORCID: 0009-0007-8558-164X)Corpus retrieval date: 28 July 2026 Contents - ICAC_bibliometric_corpus_N21.csv — the 21 documents underlying Table 1 of the manuscript (title, year, source, DOI, document type, Scopus Author Keywords, Scopus Index Keywords).- ICAC_cluster_assignment_N21.csv — per-document cluster membership from the Louvain community detection (a document may belong to more than one cluster, since its keyword set can span clusters).- ICAC_keyword_clusters_summary.csv — the five clusters, their core keywords, and document counts, as reported in Table 1. Search strategy Database: Scopus (Elsevier). Query (Title-Abstract-Keyword):TITLE-ABS-KEY( ("artificial intelligence" OR "AI adoption" OR "AI capability" OR "generative AI") AND ("absorptive capacity" OR "dynamic capabilit*") AND ("institutional theory" OR "institutional quality" OR "governance quality" OR "institutional environment")) All three concept clusters (AI, capacity, institutional/governance) were required jointly (AND), consistent with the manuscript's stated design of "combining AI-adoption, absorptive-capacity, and institutional-quality terms." Coverage window: no explicit year filter beyond what the database returned as current; resulting documents span 2022–2026, concentrated in 2025–2026. Web of Science was searched under an equivalent TS= query. All WoS documents matching the three-cluster criterion were already present in the Scopus set (same DOIs); Web of Science therefore contributed no unique documents to this corpus. This is disclosed rather than omitted. Screening No manual topical exclusions were applied beyond the three-cluster Boolean requirement itself. Two documents indexed twice in Scopus (once as a journal article, once as a duplicate conference-proceedings record of the same paper, without DOI) were deduplicated by normalized title, retaining the DOI-bearing version. Clustering method Author Keywords and Index Keywords (Scopus-assigned) were combined per document, lower-cased, and de-duplicated within document. A co-occurrence network was built where an edge connects two keywords if they co-occur in at least 2 documents (threshold = 2, matching the manuscript's stated method). Louvain community detection (python-louvain, networkx, random_state=42) was applied to the resulting weighted graph. - Nodes (keywords) at threshold 2: 22- Edges: 34- Documents with at least one keyword surviving the threshold: 14 / 21 (66.7%)- Documents below threshold (isolated keywords, no cluster assignment): 7 / 21 (33.3%) Note on revision history An earlier version of this manuscript (submitted to Innovation: Organization & Management and European Journal of Innovation Management, both declined) reported a different bibliometric result: a 33-document corpus (21 Scopus + 12 Web of Science, 2022–2026) showing two fully separated clusters and describing this as a "structural disconnect" between AI-adoption and institutional/capability-theory literature. That corpus file was not available for re-verification at the time of this revision. A fresh corpus was retrieved on 28 July 2026 using the search logic above, and re-clustered following the identical Louvain method (threshold = 2). The fresh corpus does not replicate a clean two-cluster disconnect: "artificial intelligence" and "dynamic capabilities" co-occur directly within a dominant cluster (61.9% of tagged documents). It does, however, show that no institutional-quality or governance-quality term survives as a shared keyword at the same threshold, despite all 21 documents having been retrieved specifically because they discuss such a construct in their abstract. The manuscript has been revised to report this more precise, and more modest, finding rather than the earlier claim. License This dataset is shared under CC-BY 4.0.