This study examines the dual impact of generative AI (GenAI) in knowledge work, where productivity gains risk being offset by the erosion of human expertise. By adapting the Technology-Organization-Environment (TOE) framework through qualitative interviews with professionals across healthcare, finance, and technology sectors, this thesis reveals how unguided AI adoption leads to deskilling through reduced autonomy and weakened collaboration, while structured integration preserves human capabilities. The findings of this thesis demonstrate that although GenAI automation in knowledge work in organizations enhances efficiency and productivity, overdependence undermines critical thinking and human expertise. This thesis contributes to ongoing research on AI by proposing a refined framework that identifies three key intervention areas: technological (hybrid workflows with human oversight), organizational (upskilling programs and role transformation toward strategic tasks), and environmental (ethical governance standards). Practical solutions include dynamic task redesign, skill-preservation metrics, and AI oversight protocols that balance automation with human engagement. These insights advance the human-AI collaboration discourse by providing organizations with actionable strategies to harness GenAI's benefits while maintaining workforce resilience and standards, ultimately supporting sustainable innovation in knowledge work environments.