The bright and dark sides of AI-driven transformation: evidence from Italian firms
Syed Muhammad Ali Shah, Michele Cincera, Marco CucculelliPurpose
This study examines how firm-level Artificial Intelligence (AI) capabilities are associated with innovation trajectories and firm performance, with particular attention to sustainability-oriented (green) and efficiency-oriented (brown) innovation outcomes. It further investigates how government support relates to these innovation trajectories across firms with different capability profiles and whether AI capabilities are associated with subsequent AI adoption patterns.
Design/methodology/approach
The study combines large-scale survey data from 6,732 Italian firms with LinkedIn-based information on AI-skilled employees. We construct a multidimensional AI Capability Index capturing tangible, human, and intangible knowledge sources and estimate a two-stage CDM-inspired empirical model linking capabilities, innovation outcomes, firm performance and AI adoption over a four-year horizon.
Findings
Stronger AI capabilities are associated with a higher likelihood of both brown and green innovation, although with different performance profiles over time. Brown innovation is associated with short-term efficiency and revenue gains, whereas green innovation is associated with more persistent productivity and growth effects. Government support is positively associated with green innovation among firms with stronger AI capabilities, but more closely associated with brown innovation in low AI capability settings. AI capabilities are also strongly associated with subsequent AI adoption, consistent with path-dependent capability dynamics.
Research limitations/implications
The analysis focuses on Italian firms, which may limit generalizability to other institutional contexts. AI capabilities are measured cross-sectionally, constraining direct observation of capability evolution over time. Although LinkedIn data enables longitudinal tracking of AI adoption, adoption may be undercounted for micro firms or firms with limited online presence. Future research using panel data and cross-country settings could further strengthen causal inference and external validity.
Practical implications
For managers, the results highlight the importance of developing balanced AI capabilities rather than focusing solely on technology acquisition. Firms with stronger capabilities appear better positioned to align AI-related innovation with long-term sustainability and growth objectives. For policymakers, the findings suggest that broad AI support mechanisms may be associated with different innovation trajectories depending on firms' underlying capabilities, underscoring the importance of capability-sensitive policy design.
Social implications
By showing that AI capabilities are associated with different innovation orientations, the study highlights the societal importance of capability development in shaping the broader implications of AI adoption. Policy support aimed at strengthening firms' absorptive and organizational capacity may help align AI-enabled innovation with environmental and societal objectives, reducing the risk that public support becomes associated primarily with short-term efficiency-oriented innovation paths.
Originality/value
The study contributes to research on AI and innovation by introducing a multidimensional measure of AI capabilities and by empirically examining how firm capabilities and public support are jointly associated with AI-enabled innovation. Conceptually, the study develops the Triple Helix Twin as a capability-contingent analytical lens for understanding heterogeneous innovation trajectories under shared institutional conditions. The findings contribute to debates on AI adoption, industrial policy, sustainable innovation and the organizational implications of AI-driven transformation.