The paradox of human–AI symbiosis: policy intent, organisational adoption, and the limits of responsible scaling

Human–AI symbiosis is frequently promoted as a strategic imperative for organisations, yet its implementation remains largely confined to low-risk and experimental contexts, even where strong policy support exists. This study investigates why the transition from policy intent to organisational adoption remains difficult. Drawing on interviews with 20 senior managers operating in an advanced AI governance environment, we identify an asymmetry between institutional ambition and organisational practice. While institutional actors generate strong normative pressure to adopt responsible AI, they remain limited in implementation. Meanwhile, organisations adopt AI technologies more rapidly than governance mechanisms evolve to support responsible scaling. This misalignment produces six challenges: infrastructural constraints, tensions between speed and deliberation, bias amplification, limited transparency, regulatory complexity, and competency gaps. In response, organisations develop three adaptive practices: hybrid decision architectures, explainability procedures, and iterative system design. Together, these practices constitute a governance-through-practice logic that enables managers to bridge the gap between policy aspirations and operational realities. By offering a processual account of human–AI symbiosis, the study shows that responsible scaling is slower, more contested, and more practice dependent than research suggests. It also advances AI governance scholarship by illuminating the work required to translate normative commitments into managerial routines.