Putting an AI agent in front of patients without an incident
Scope boundaries, acuity escalation, the never-answer list, logging and sign-off: the controls that have to exist before a model speaks to a patient.
conversational ai
Containment vs resolution: the chatbot metric that misled us
A hospital chatbot reporting 80 percent containment may be resolving less than half of what patients wanted. Why the default metric misleads, and what to measure instead.
