<![CDATA[Artificial intelligence (AI) has intensified debates surrounding the contemporary productivity paradox, where rapid technological progress coexists with uneven improvements in measured productivity. Although growing evidence shows that AI can significantly enhance task-level performance—reducing completion time, improving output quality, and standardizing decision processes—these gains do not always translate into consistent firm-level productivity outcomes, particularly among small and medium-sized enterprises (SMEs) operating in platform-mediated digital markets. This article develops a conceptual framework that revisits the AI productivity paradox through a multi-level theoretical perspective. Integrating insights from productivity paradox research, general-purpose technology theory, task-based technological change, and platform ecosystem scholarship, the study proposes that AI-induced productivity gains propagate unevenly across four analytical layers: tasks, SMEs, platforms, and digital ecosystems. Three generative mechanisms—complement lag, measurement wedge, and compounding learning effects—explain how productivity gains are translated, partially observed, or redistributed across these levels. While SMEs may experience delayed or weakly measured productivity improvements due to complement constraints and measurement limitations, platform infrastructures can accumulate accelerated gains through data-enabled learning and cross-merchant aggregation. The framework introduces productivity divergence as a concept explaining how ecosystem-level efficiency can increase even when individual firms experience uneven productivity outcomes.]]>
Copyrights © 2026
