Across the EU, analytics capability — not AI adoption alone — is linked to shifts in middle-management structures; business-intelligence diffusion correlates with managerial reconfiguration while broad AI uptake shows weaker direct association.

Main Finding

AI adoption across EU firms is associated with organizational change, but not primarily through simple managerial displacement. Instead, changes in middle-management structures are most strongly linked to firms’ analytical capabilities—especially the diffusion of Business Intelligence (BI) software—and occur mainly in more digitally advanced organizational environments. AI intensity alone does not predict reductions in middle-management employment at the country level.

Key Points

• Conceptual framing: internal enterprise capabilities (digital + innovation) → AI adoption mechanisms (general use; administrative/management use) → organizational restructuring (centralization, hierarchical change, labor reallocation, role of middle management).
• Heterogeneous landscape: substantial cross-country variation in digital integration, innovation practices, and AI adoption intensity across EU members.
• BI (data-analytics capability) matters: higher BI software diffusion is positively associated with measurable changes in middle-management employment shares in digitally advanced contexts.
• ERP (enterprise integration) and organizational/managerial innovation are treated as enabling conditions but BI shows the clearest link to managerial-architecture change in the analyses.
• AI use per se (share of firms using at least one AI technology) is not sufficient to detect managerial displacement in aggregate country-level data; the organizational consequences depend on capability configurations and the contexts of AI integration (e.g., AI used in management/administration).
• The study shifts focus from adoption determinants to organizational implications and provides an indirect, country-level quantitative assessment of AI-driven restructuring within the EU.

Data & Methods

• Data sources: Eurostat comparative datasets (enterprise-level indicators aggregated to country level). The paper references the 2025 Eurostat enterprise AI survey (157,000 firms sampled of ~1.53 million enterprises in the EU) and other Eurostat indicators (ERP use, BI use, business process and managerial innovation).
• Outcome variable: share of employment in middle-management occupations (ISCO-08 groups 12–13).
• Key explanatory variables: share of enterprises using ERP; share using BI software; share introducing business-process innovations; share introducing managerial/HRM innovations; share of enterprises using any AI; share of AI-adopting enterprises using AI for business administration/management.
• Empirical approach:
  • Stage 1: K-means cluster analysis to identify relatively homogeneous groups of EU countries based on digital and innovation capabilities and AI adoption mechanisms.
  • Stage 2: Linear regression analysis (country-level) to estimate associations between capability/AI indicators and middle-management employment shares, with attention to interactions/conditional relationships in more digitally advanced clusters.
• Interpretation cautions noted by authors: indicators are indirect proxies for organizational restructuring; analysis is comparative and associative (country-level aggregation), limiting causal claims and firm-level inference.

Implications for AI Economics

• Organizational capital and analytics investments shape AI’s labor effects: economic models of AI adoption should incorporate complementary investments in analytics (BI) and ERP systems, as these condition how AI translates into role reconfiguration rather than simple labor substitution.
• Heterogeneous, contextual effects: cross-country and cross-firm variation in digital and innovation capabilities implies that aggregate labor-market predictions (e.g., manager job losses) will be misleading without accounting for capability complementarities and organizational design.
• Redistribution of managerial tasks, not uniform displacement: AI may reallocate managerial labor toward interpretive, governance, and hybrid coordination roles (centralized model governance with decentralized execution), affecting demand for different managerial skills and potentially wages and productivity in non-linear ways.
• Policy and regulation: EU AI governance (AI Act and related plans) interacts with firm-level restructuring—regulatory emphasis on risk management, transparency and governance may alter firms’ incentives to centralize control or invest in BI/ERP, with implications for competition, accountability, and labor protection.
• Research priorities for AI economics: need for firm-level longitudinal and causal studies to estimate productivity gains, wage effects, and the distributional impacts of AI-enabled restructuring; evaluation of returns to investments in BI/ERP and managerial retraining; and modeling of how algorithmic centralization affects firm boundaries, market power, and labor bargaining.

Limitations noted by the study (relevant for economic interpretation): reliance on country-aggregated Eurostat proxies (indirect measures), cross-sectional or limited-time aggregation limiting causal inference, and sample composition skewed toward small firms in the Eurostat survey. Future work should use firm-level panel data, richer measures of managerial tasks and wages, and exploit quasi-experimental variation in AI or policy exposure.