What an AI receptionist does all day (and what it never does)

At 7:52 on a Tuesday morning, before anyone has unlocked the front door, a new patient calls to ask whether we treat thyroid problems and how soon she could be seen. She gets an answer in the practice’s own words, two openings with the right practitioner, and a text with the intake form before she has finished her coffee. Nobody at the practice has done anything yet.

That is the unglamorous truth about an AI receptionist: most of what it does is small. But the small things are exactly what a two-person front desk cannot keep up with, and they are where patients quietly give up on you.

The morning: confirmations and the reschedule dance

The first hour is confirmations. Every visit within 48 hours gets a text; replies of “yes”, “C” or a thumbs-up are read as confirmations, anything else is read for meaning. A patient who writes “can we do 3 instead?” is offered the nearest open 3 o’clock, and the calendar updates itself. A patient who does not reply by mid-morning gets a call. If the call reaches voicemail, the visit is flagged for a person, and it shows up in the schedule as not reached rather than silently remaining on the books.

Two things matter here. The receptionist only offers times that are actually open for that practitioner and that service, because it books from the same list the practice manages under Services & pricing. And every move it makes is written into the appointment, so when the front desk looks at the day, the story is already there.

Midday: new patients and the questions behind the questions

New-patient calls are rarely about scheduling alone. “Do you take insurance?” often means “will this cost more than I expect?” The receptionist answers what it has been approved to answer, from the practice’s knowledge base: membership pricing, what a first visit includes, how superbills work. It collects the details the practice needs for a first visit, offers the consultation, and sends the forms. What it does not do is guess. If the answer is not in the knowledge base, it says so and offers a callback, with the question written down for the person who returns the call.

A good receptionist is not the one who knows everything. It is the one who knows exactly what it is allowed to say, and who to hand the rest to.

The line it does not cross

A patient on a new medication calls to say she has felt dizzy since Monday. The receptionist does not reassure her and it does not diagnose. It acknowledges, tells her to call 911 if it gets worse or she faints, books the soonest visit with her practitioner, and routes the conversation to me with the full transcript and a drafted reply. That routing rule is not a setting a practice can switch off. Clinical questions go to a clinician. Every time.

Afternoon: refills, results and “where do I park?”

Refill requests are routed to the prescriber with the pharmacy already attached. Patients asking about results are told when their practitioner will review them, never what the numbers mean. Parking, hours, telehealth links, what to bring: the receptionist handles the whole long tail of practical questions that used to interrupt the front desk twenty times a day, and it handles them in the practice’s voice, not a generic one.

After hours

At 6:15 PM the phones do not go to voicemail. They go to the receptionist, which books, confirms and answers approved questions overnight, and leaves a tidy queue for the morning: three visits to confirm, one callback requested, one refill for the prescriber. The practice opens to a list, not a backlog.

What the front desk does instead

The people are still there. They now spend their time on the calls that need a person: a nervous patient before a first visit, a billing question with history behind it, the patient who prefers a human and says so. The AI took the volume; it did not take the relationship.