AI Receptionist for Semble Private Practices

A private patient rings a consultant's rooms with three questions, usually in this order: can I see Mr Adeyemi, what will it cost, and will my insurer pay. A medical secretary answers those in her sleep. A generic voice agent answers the first, fumbles the second, and has no idea about the third, which is exactly why most of them are not trusted with a private line. Getting all three right, on every call, is the bar a Semble practice actually needs cleared.

We build the Semble connection, and this piece is about clearing that bar: the three questions, and what happens underneath when the answer becomes a booking.

The order matters, because a caller who cannot get a straight answer on cost or cover will not book, however pleasant the conversation. 2care answers all three from what the practice holds in Semble, not from a script that guesses.

The question	Where the answer comes from	What the caller gets
Which consultant, and when	Live consultant availability in Semble	A real slot with the right consultant
What will it cost	The self-pay fee the practice has set	The price, before they commit
Will my insurer pay	The cover rules the practice configures	A clear yes, no, or check with the insurer

The third answer is the one generic agents skip and the one that decides the booking. Where cover genuinely cannot be confirmed on the call, the agent says so plainly and routes the caller rather than guessing, because a wrong answer on cover is worse than no answer: it produces a patient who arrives expecting to be covered and a consultant's office left to have an awkward conversation.

Take a real sequence. "I'd like to see a cardiologist about some palpitations, and I'm with Bupa." The agent recognises a new private enquiry, a described symptom that is a booking reason rather than a clinical emergency, and a named insurer. It offers the cardiologist's next suitable slot, states the self-pay fee in case the patient prefers it, confirms whether Bupa is recognised for that consultation, and tells the patient what the insurer will want, an authorisation code or a referral, before the appointment. What the patient hears in ninety seconds is what a good secretary would have told them, and the booking is in Semble before they hang up.

How a booking reaches Semble, and how fast

Underneath the three answers is one mechanism. 2care connects to Semble through its GraphQL API, authorised with a scoped credential set during onboarding, with nothing installed at the practice. Availability is queried live per consultant and appointment type, so an offered slot genuinely exists, and the booking is written against the resolved patient and validated by Semble against its own rules, so a slot taken mid-call fails cleanly rather than being assumed free.

The speed is what makes it usable on a private line, where the caller expects the polish of a good secretary. The agent replies in about 480 milliseconds, ninety-fifth percentile under 700, on a native voice pipeline rather than a wrapper over a third-party service, so querying Semble live happens inside a conversational beat rather than as an awkward pause. The platform holds 99.9% uptime and takes 1000+ concurrent calls, and the write carries an idempotency key so a network stutter never books the same consultation twice. Turn-taking waits for a finished sentence, so an older patient reading out a policy number is not cut off halfway through it.

Consultant-specific booking, without the secretary in the middle

A private practice is really a set of independent diaries, one per consultant, each with its own rules: the appointment types they offer, the length of a first consultation versus a follow-up, the sessions they run at which location. A booking that ignores those is a booking a secretary has to move.

2care books into the right consultant's diary under that consultant's own rules, reading the live Semble availability and honouring the appointment types and durations the practice has set. A caller who describes a problem rather than naming a consultant is matched to the right one from the practice's configuration, and where a case genuinely needs a human judgement about which consultant should see it, the call is routed rather than forced into a slot. The point is not to replace the secretary's judgement but to take the routine bookings off her desk so the calls that need her get her.

There is a subtler benefit for a multi-consultant practice. When several consultants share a front desk, a caller who reaches a voicemail does not know which consultant's secretary they have missed, and the enquiry can sit in the wrong inbox for a day. Reading the live Semble diaries directly, the agent routes the enquiry to the right consultant's book at the moment of the call, so nothing lands in a shared voicemail to be sorted later. On a chambers or a group of consultants sharing rooms, that single change removes a whole category of lost enquiry.

Fees and cover, answered the same way every time

The reason cost and cover are hard is that they are inconsistent when a human answers them under pressure, and inconsistency on money erodes trust faster than almost anything else on a private line. Answered from the fees and cover rules held in the practice's setup, the reply is the same every time and correct every time: the self-pay fee for a first consultation, whether a given insurer is recognised, and what the patient needs to have ready.

Where the honest answer is "your insurer needs to confirm that," the agent says exactly that and captures the details so the practice can follow up, rather than inventing a reassurance it cannot stand behind. A caller who books knowing the fee and the cover position is a caller who turns up, and a practice that never has the fee conversation at reception is a practice with a calmer front of house.

The cover answer is also where a practice protects its own time. An unrecognised insurer discovered at reception means a patient turned away or a bill written off, both of which cost more than the consultation. Answered on the booking call, an insurer the practice does not accept becomes a choice offered up front, self-pay or find a covered consultant, rather than a surprise on the day. The agent captures whatever the insurer will need so the practice is not chasing an authorisation code the morning of the appointment, which is the small administrative friction that quietly loses private practices their afternoons.

The record and the rules the UK sets

Before anything is written, the caller is matched to their existing Semble record and confirmed on a date of birth, so a returning patient keeps one file; a genuinely ambiguous match goes to a person rather than a guess, using patient matching, which runs before a slot is ever held, so an error never reaches the consultant's diary.

Because a booking builds a transcript, matches a patient and writes to Semble, it processes personal data rather than passing it through, which makes the vendor a data processor under UK rules and, for any US work, a business associate under HHS guidance. So the data is held in region under a data processing agreement, encrypted in transit and at rest, never used to train the models behind the conversation, and every write is reversible. The compliance page sets out the detail.

What 2care changes for a consultant's rooms

A consultant's rooms are not short of a booking system; Semble is a good one. They are short of a phone that answers with a secretary's competence when the secretary is on another call, at lunch, or gone home for the evening. 2care is that phone. It answers each call on the first ring in about 2 seconds, quotes the fee, checks the cover, books the right consultant and writes it into Semble, without a person keying anything in.

What a practice feels is not a faster booking on the calls the secretary already took. It is the enquiries that used to reach an answering machine at 7pm now booked, with fee and cover already settled, by morning, and the secretary's day cleared of the routine so the calls that genuinely need her judgement actually reach her. 2care does not replace the secretary; it hands back the third of her day the phone used to eat, and it never puts a caller on hold to do it.

Where this is right for a private practice, and getting live

This is right for a practice whose bookings turn on consultant, fee and cover, that keeps those three current in Semble, and that has a secretary or clinician to take the calls needing a human judgement. Given those, the benefit lands from the first day the line is live, and the busiest weeks, the ones a single secretary cannot cover, are exactly when it earns its place. A practice whose fees vary case by case in ways only a person can weigh will still route more of those calls, which is a fair thing to know going in, and the agent handles it by asking rather than inventing.

Getting there is three moves over weeks. The connection is authorised to read availability and write bookings through the Semble GraphQL API. The setup is mapped: consultants, appointment types and durations, self-pay fees, recognised insurers and cover rules, and the route for a call needing a human. Then the line is tested against a sandbox with the secretary listening, before it is switched over and watched while the fee and cover answers settle against real calls.

Frequently asked questions

Can it quote a self-pay fee on the call?

Yes, from the fees the practice has set in its configuration, so the quote is the same every time and correct. Where a fee genuinely depends on something only a person can judge, it says so and routes the call rather than guessing at a number.

How does it handle an insurer question?

It answers from the cover rules the practice configures: whether an insurer is recognised and what the patient needs ready. Where cover cannot be confirmed on the call, it says so plainly and captures the details for the practice to follow up, rather than promising cover it cannot verify.

Does it book into the right consultant's diary?

Yes. It reads each consultant's live availability in Semble and books under that consultant's own appointment types and durations. A caller who describes a problem rather than naming a consultant is matched to the right one, and a genuinely finely-balanced case is routed to a person.

How is our data handled?

It is held in region under a data processing agreement, encrypted in transit and at rest, and never used to train shared models. A returning patient is resolved to one Semble record, and an ambiguous match goes to a person rather than creating a duplicate file.

Does it work outside the secretary's hours?

Yes, every hour including evenings and weekends, at any call volume, with no busy signal. The enquiries that used to reach an answering machine are answered, quoted and booked, and anything needing a human is captured and routed for the morning.

Hear it answer fee and cover, then book

The call worth judging on a Semble line is the real private-practice one: which consultant, what it costs, and is it covered, answered cleanly and turned into a booking with the right consultant at the right fee.

See how the same connection runs on other systems across integrations, or hear a private consultation booked, fees and all, on a live call when you book a demo.