Abstract
Digital transformation is a key force driving high-quality economic development, yet the economic consequences of different digital technology directions vary significantly. Existing research often treats it as a homogeneous whole, overlooking the "diverse direction" characteristics of technologies, making it difficult to explain the "productivity paradox." From the perspective of dynamic capability reconstruction, this paper constructs a theoretical model of "Digital 'Diverse Direction' Transformation-Dynamic Capability-Total Factor Productivity (TFP)." Using the LDA topic model to conduct semantic analysis on the annual reports of A-share listed companies, we categorize digital transformation into five directions: Artificial Intelligence (AI), Big Data, Cloud Computing, Blockchain, and Digital Technology Application. We empirically test their differentiated impacts on enterprise TFP and the underlying mechanisms. The findings are as follows: First, while digital transformation overall significantly promotes enterprise TFP, there is significant "technological heterogeneity" among different directions. AI has the strongest promoting effect, followed by Big Data and Digital Technology Application, whereas the impacts of Cloud Computing and Blockchain are not yet significant. This provides micro-evidence for explaining the "productivity paradox." Second, mechanism tests indicate that dynamic capability reconstruction is the key transmission path; digital transformation enhances TFP by strengthening organizational coordination and integration capabilities, change and reconstruction capabilities, and learning and absorption capabilities. Third, heterogeneity analysis reveals that digital transformation exerts a more pronounced effect on enhancing productivity in labor-intensive and technology-intensive industries, as well as in the manufacturing sector. These conclusions deepen the theoretical understanding of the economic consequences of digital transformation and provide empirical evidence and decision-making references for enterprises to choose adaptable and differentiated transformation paths.