How should clinical work be organized once AI systems take part in it? My research studies this question empirically, across three connected directions: verifiable AI for clinical decisions, AI support that carries information across the care pathway, and embodied and immersive AI in patient care.

AI systems are likely to become involved in more stages of clinical work over the coming years. They may help draft and critique plans, follow a patient through the care pathway, support handovers between clinicians, and reach clinicians through new interfaces. Some of this will be driven by technical progress alone. The design questions will not be settled that way. What the clinician’s work becomes, how information moves safely through a care pathway that includes machines, which interfaces help rather than distract, and how patients experience care in which intelligence is distributed across people and systems are all open questions. I want to answer them with evidence while the answers can still influence how these systems are built and used.

All systems I develop are advisory. They are evaluated in simulation first, then in retrospective evaluation, and only after that in any prospective use. None of them places orders or replaces the judgment of a clinician.