This week's developments connect EHR context, home monitoring and language models. Each promises to make AI more responsive to the person and situation rather than producing a generic answer.
What practices should take from this
Context should be purposeful. Give a system the minimum information needed for the approved task, show users what informed the result and prevent one workflow's data from silently becoming another workflow's input.
For patient-facing use, context also includes language, access, urgency and the channel through which a response will arrive. Technology that ignores those realities can widen a gap while appearing personalized.
The design standard is not maximum data. It is the right context, available to the right workflow, with evidence, boundaries and a human path when the system is uncertain.