A Study on the Types and Structural Mechanisms of AI-based Technology Commercialization among Korean SMEs

This study empirically examines the status and effectiveness of AI-based technology commercialization among Korean small and medium-sized enterprises (SMEs) following the diffusion of generative AI, and investigates how AI adoption shapes entrepreneurial performance and industrial innovation structures, aiming to derive national-level policy implications.To this end, the study conducted in-depth interviews with ten domestic SMEs in the manufacturing, medical/healthcare, and service (software) sectors, complemented by a comparative analysis of AI commercialization support policies in the United States, China, and Japan, as well as domestic policy documents.The study developed a "Market/Technology Novelty Matrix," integrating the logic of the Ansoff Matrix and the BCG Matrix, to classify the innovation positioning of the surveyed firms.The analysis identifies four distinct patterns of AI utilization-operational-efficiency-oriented, product-enhancement-oriented, new-business-creation-oriented, and exploration/preparation-oriented-and shows that performance differences stem less from the sophistication of the technology itself than from the availability of causal data assets and the maturity of industry-specific supplier ecosystems.Based on these findings, the study proposes a "Six-Stage AI Innovation Mechanism"-spanning external environmental pressure, strategic decision-making, core input factors, differentiation of AI utilization types, commercialization outcomes, and structural barriers-together with an integrated "AI Adoption-Policy Linkage Framework" that connects this mechanism to government policy.The findings suggest that the success of AI-based technology commercialization depends less on algorithmic sophistication than on non-technical factors such as data asset maturity, organizational absorptive capacity, and supplychain structure.Accordingly, government policy should move beyond uniform technology-diffusion approaches toward firm-type-tailored support, early-stage causal-data infrastructure development, industry-academia-research matching ecosystems, and a shift toward execution-based performance indicators.