Abstract
AI research on work has largely emphasised automation, augmentation, bias and decision support. An emerging psychological phenomenon remains under-theorised: employees can internalise AI systems as extensions of their professional self, producing human-AI identity blending that alters competence signals, autonomy and leadership legitimacy. This conceptual paper synthesises peer-reviewed research published primarily between 2023 and 2025 and integrates social identity theory, identity work, self-determination theory, sensemaking and extended mind perspectives to develop a multi-level framework. We define human-AI identity blending as a state in which workers experience AI systems as cognitively, professionally and symbolically integrated into their self-concept at work. We explain how leadership framing, boundary setting and accountability design shape whether blending becomes healthy identity integration (supporting learning, calibrated trust and psychological safety) or unhealthy overdependency (producing skill atrophy, authority erosion and ethical risks). The paper contributes by (1) clarifying the construct and its dimensions, (2) theorising leadership dilemmas when competence is co-attributed to algorithms, and (3) proposing testable propositions and a future research agenda. Practical implications outline actionable leadership behaviours for hybrid teams, including transparency routines, learning-oriented coaching, and governance practices that preserve human agency while leveraging AI capabilities.