Most health AI so far has been shaped like a conversation: you ask, it answers, the exchange ends. That shape is a poor fit for clinical work, where the burden is not answering questions but doing the work that surrounds the questions.
The difference that matters
A chatbot is reactive and stateless. An agent holds context across time and acts without being prompted each step. In a clinic that distinction is the whole value: the note that is already written when the patient leaves, the letter already drafted, the previous report already read and compared before you open it.
What an agent does that a chat window cannot
- Carries the thread between visits. What was ordered, what came back, what is still outstanding.
- Produces work rather than answers. A referral letter in the right format, not a description of how to write one.
- Acts in the background. The reading happens before the consultation rather than during it.
- Stops at the signature. It drafts; the clinician composes and signs. Nothing is issued without that.
The part nobody mentions: it costs more
An agent that runs through an entire consultation consumes far more compute than a chatbot answering one question. That has a direct commercial consequence — a per-seat price set by intuition can easily cost more to serve than it earns, and the only way to know is to measure real consultations rather than model them.
Anyone selling you an agent at a chatbot price has either solved something remarkable or has not done that arithmetic.
This article explains how our technology works. It is health information, not medical advice.