Despite the promising potential of artificial intelligence (AI) in mental healthcare, practitioners and researchers consistently identify resistance to adoption as a critical bottleneck in implementing these technologies. This resistance manifests across the mental healthcare ecosystem, with both practitioners and clients expressing varying degrees of hesitation to integrate AI-based solutions. Prior research has framed this phenomenon as algorithm aversion – a preference for human over AI decision-making, even when AI demonstrates superior performance. However, in this talk, I argue that current research has failed to fully explain AI adoption barriers due to two critical gaps: the absence of a robust theoretical framework and insufficient attention to the social-relational contexts in which AI systems are deployed. The first part of this talk demonstrates how the lack of a theoretical framework has led researchers to misidentify or overemphasize certain adoption barriers while overlooking others. The second part presents a novel framework for understanding AI adoption through a social-relational lens, offering a more nuanced approach to studying and addressing implementation challenges in mental healthcare settings.