Policy debates often assume artificial intelligence (AI) will improve both healthcare efficiency and doctor–patient relationships, whereas a critical strand of the bioethics literature expects gains mostly in efficiency. Little is known about what happens when an AI system is integrated into care pathways and the values that clinicians and developers believe should guide the use of AI in healthcare. This study aims to (1) understand experiences of clinicians using AI in practice, (2) understand the values clinicians and developers prioritise in the deployment of AI in healthcare and (3) consider the ethical question of how the doctor–patient relationship can be positioned within healthcare systems focused on increasing productivity. Semi-structured interviews were conducted with clinicians using AI in their practice (n=11) and AI developers (n=10). Interviews were transcribed, thematically coded and analysed. Clinicians and developers expressed diverse views on what should be optimised following AI implementation. Developers prioritised system-level outcomes associated with efficiency, such as patient throughput and error reduction. In addition, clinicians highlighted the importance of prioritising the doctor–patient relationship. The absence of a shared understanding of what aspects of healthcare should be optimised creates an ethical blind spot. While aspects of care traditionally associated with efficiency are often seen as incompatible with the relational dimensions of care, this paper suggests that there may be benefits in conceptualising efficiency more broadly than current policy framings allow. However, whether such a reframing translates into practice depends on the organisational incentives and institutional structures within which AI is deployed as opposed to the technology alone. Without attending to these, the integration of AI in healthcare risks entrenching a narrow conception of efficiency that prioritises process metrics over patient-relevant outcomes.