As generative artificial intelligence (GenAI) solutions proliferate, organizations face a critical challenge: how to embed these broadly capable tools into value-creating routines. This study reveals a more nuanced reality of how GenAI solutions are embedded across small and large organizations. Drawing on 19 in-depth interviews with a small e-commerce startup and a large biotech incumbent, we identify how organizational scale and decision-making structures, associated control space, and the decision latitude afforded to individual actors, jointly shape GenAI integration. While the smaller firm rapidly experiments with and deploys large language models (LLMs) to bolster productivity, the larger incumbent struggles to leverage these innovations meaningfully. We theorize these divergent outcomes through the lens of decision incongruence, wherein formal commitments to GenAI adoption remain misaligned with frontline practices. This misalignment arises not from resource deficits but from entrenched routines, opaque rationales, and fragmented authority, which collectively constrain the effective enactment of new technological capabilities. Our findings extend organizational design and technology adoption literature by foregrounding the structural and relational underpinnings of GenAI integration. In doing so, we shed light on how the interplay of decision-making autonomy, hierarchical rigidity, and interpretive frames either enables or impedes the capacity to transform lofty aspirations for GenAI into enduring sources of organizational value.
