This Evidence Note examines why current implementation science frameworks fail to adequately address clinical AI deployment. While existing models assume stable interventions, clinical AI operates within dynamic systems characterized by continuous change. The note introduces Dual Drift as a central mechanism: the simultaneous evolution of biological systems and AI models. This interaction creates a structural mismatch between static implementation frameworks and dynamic clinical reality. Model drift is reframed not only as a technical issue, but as a system-level signal that may reflect underlying instability. Without interpreting drift in relation to system state and trajectory, interventions risk reinforcing incorrect interpretations rather than preventing failure. The analysis connects implementation science with dynamic systems theory within the Universal Resonance Model (URM), emphasizing the importance of state awareness, trajectory, and early instability signals in clinical AI governance.
Anita Domargård (Sat,) studied this question.