Healthcare is a complex sociotechnical system, and while current systems better integrate with existing technology, they struggle to integrate with workflow. Using reflection-in-action as a lens, we analyze existing approaches and develop a new perspective on the design of healthcare technology. Clinicians are constantly reflecting in-action; they ask questions or perform an exam and the situation “talks” back to them and informs the diagnosis. We aim to integrate this interplay between cognitive tasks and feedback from the environment into healthcare technology. Current cognitive support systems talk back at the wrong time; they don’t know enough about what is going on and therefore they are often wrong. Alert fatigue and non-use are the result. We argue that reflection-in-action has to be explicitly designed for by embedding cognitive assistance not into a single technology, but by focusing on human-computer teamwork. A successful cognitive assistant facilitates reflection-in-action by challenging the clinician with an alternative perspective throughout the day– even in front of the patient– thus encouraging clinicians to articulate their thinking and enabling meaningful cognitive support.