ContactBook a call

Industries/ Healthcare/ Front desk assistants

An AI receptionist for the routine questions, and nothing clinical

It answers opening hours, prices, parking and what to bring on your website and messaging channels, and books from your real diary once a patient is verified. Anything clinical or personal goes straight to a person, with the conversation attached.

A calm clinic reception where the receptionist greets a patient in person while a monitor beside her shows an online chat being handled, waiting area with plants.
Used by
Patients and enquirers, with reception taking every handover
Takes the place of
Repeat calls about hours, prices, parking and what to bring
Connects to
Your diary, website, WhatsApp, SMS and the reception inbox
Built with
Our ai receptionists & chat service

Where it hurts today

A normal week, before it exists

Most of the calls reception takes are the same dozen questions. What time do you open on Saturday? Where do I park? Do I need a referral? Do you work with my insurer? What should I wear for physiotherapy? How much is an initial assessment? Each one takes a minute, every one interrupts the person standing at the desk, and the answers are already written somewhere on your website.

After hours the questions keep coming, through the contact form, the voicemail and the clinic mobile on WhatsApp. They are answered the next morning, in a batch, by whoever gets to them first. A patient who wanted to move tomorrow's 8am appointment finds out at 8.30am that the message was read too late, and the slot goes unused while someone else waits for one.

The quieter risk is in how those messages get answered. A busy receptionist replying on WhatsApp can be drawn into a question about pain after treatment, or confirm an appointment to someone who has not proved who they are. Nobody intends it. It happens because the channel was never designed for the job, and nothing records what was said or by whom.

What it does

Front desk assistants, feature by feature

An assistant that does the routine work well and knows exactly where its job ends.

  1. 01

    Answers from your own approved information

    It answers from an answers library your team writes and approves: opening hours, sites, parking, prices, what to bring, which insurers you work with and your cancellation terms. It does not answer from general knowledge. When the library does not cover a question, it says so and passes it to a person.

  2. 02

    A firm clinical boundary

    It never gives advice, interprets symptoms, comments on medication or results, or suggests which treatment someone needs. When a message touches any of these, it stops, shows the urgent help wording your clinical lead has written, and hands the conversation to reception. That boundary is tested before launch and after every change.

  3. 03

    Appointment booking against the real diary

    Once a patient has proved who they are, with a code sent to the mobile number on your records, the assistant can book, move or cancel appointments of the types you allow. It reads availability from your practice system, so it offers the same slots reception would see.

  4. 04

    Handover to reception with the whole conversation

    When it hands over, reception receives the full conversation, whether the patient was verified and what they asked for, in the inbox or helpdesk they already use. The patient is told when someone will reply, based on your opening hours, so nobody is left wondering whether the message was seen.

  5. 05

    Nothing personal without verification

    Until someone is verified, the assistant does not confirm whether they are a patient, discuss any appointment or repeat any detail from the record. Parents and carers are handled by the rules you set, and anything outside those rules goes to a person rather than being guessed at.

  6. 06

    Every conversation logged and reviewable

    Every conversation is logged, with what the assistant said and why it handed over. Each week reception and the practice manager see the questions it could not answer and the handovers it made, add answers to the library where they belong, and approve each change before it goes live.

A day with it

How the day runs instead

A normal day at a physiotherapy clinic, with the assistant on the website and WhatsApp and reception taking the handovers.

  1. 6.50am
    Patient

    Before work, a patient asks on WhatsApp whether the clinic opens on Saturday and where to park at the East site. The assistant says it is an automated assistant, answers from the answers library and links to the directions page. Nobody at the clinic needs to see it.

  2. 7.20am
    Patient

    Another patient wants to move Thursday's appointment. The assistant sends a code to the mobile on her record, she types it in, and it offers three times with the same physiotherapist. She picks Friday at 9.40am, and the change is in the diary before reception arrives.

  3. 9.05am
    Receptionist

    He opens the overnight handovers. One patient mentioned swelling after yesterday's treatment. The assistant stopped, showed the urgent help wording and passed the conversation on. He follows the clinic's own procedure and puts it in front of the treating clinician, who rings the patient that morning.

  4. 1.30pm
    Practice manager

    She reviews the questions the assistant could not answer this week. Several asked whether a particular insurer is accepted, which is not in the library. She writes the answer, reception checks it, and she approves it, so from tomorrow the assistant can answer it.

  5. Friday 4.00pm
    Clinical lead

    He reads every handover marked clinical from the week and checks the assistant stopped at the first mention each time. One patient described a symptom in an unusual phrase, so he asks for that phrase to be added to the stop rules and the tests.

What it connects to

Works with what you already run

The assistant can only do what its connections allow, so each one is chosen deliberately and limited to what the job needs.

  • Practice management system

    The assistant reads availability and appointment types, and can book, move or cancel for a verified patient. It has no access to clinical notes, results or letters, and every action it takes is logged.

  • Website and messaging channels

    Chat on your website, plus WhatsApp through the WhatsApp Business Platform and SMS where you use them. One answers library and one set of rules sit behind every channel, so patients get the same answer wherever they ask.

  • Reception inbox or helpdesk

    Handovers land in the shared inbox or helpdesk reception already works from, with the conversation, the patient's verification status and the reason for the handover attached, so nobody has to ask the patient to repeat themselves.

  • AI model provider

    The language model is chosen with you, on terms that stop your conversations being retained or used for training, hosted in the region your obligations require, or run inside your own cloud account.

The first version

Start small, then grow it

A first version should be useful on day one and small enough to get right. This is where we would usually draw the line, and what tends to follow once people are using it.

In the first version

  • Website chat answering hours, sites, parking, prices and preparation questions
  • An answers library your team writes, reviews and approves
  • Clinical stop rules with urgent help wording written by your clinical lead
  • Handover to the reception inbox with the full conversation attached
  • A log of every conversation and a weekly review of handovers

Once it is in use

  • Booking, moving and cancelling appointments once identity has been verified
  • WhatsApp and SMS alongside website chat, from the same answers library
  • Answers in the languages your patients write in, with safety wording translated by people
  • A voice line for opening hours and directions only, if it suits your callers

Buyer's guide

Before you commission one

Four decisions worth making before you put an AI receptionist in front of patients.

  1. 01

    A packaged AI receptionist or a custom one

    There are packaged AI receptionist and chatbot products for clinics, and some practice software includes one. If a product answers from your own approved content, hands over to a person cleanly and stops on anything clinical, it may be enough. Custom earns its place when you need firm, testable clinical stop rules, booking against your real diary, your choice of model and region, and logs you control.

  2. 02

    What to gather first

    Collect the questions reception answers most, from the phone log, the contact form and the clinic's messaging inbox. Gather the approved answers that already exist on your website and in your policies, and list the topics the assistant must never touch. Ask your clinical lead to draft the urgent help wording early, because nothing launches without it.

  3. 03

    Who should own it

    The practice manager should own the answers library and the weekly review. Your clinical lead should own the stop rules and the urgent help wording, with the right to pause the assistant at any time. Your data protection lead owns where conversations go and how long they are kept, and reception owns the handover inbox and should be consulted on every change.

  4. 04

    How you will know it is working

    Watch what reception notices: fewer repeat calls about hours, parking and prices, handovers that arrive with everything needed to act, and patients who do not have to repeat themselves. Read the handovers and the unanswered questions every week. Above all, check that every message mentioning symptoms or medication was handed over, because that is the one result that must never slip.

Terms worth knowing

The words you will meet

Software as a medical device
Regulators such as the FDA in the US and the MHRA in the UK use this idea for software intended for a medical purpose. Software that informs diagnosis or treatment can fall into it, with clinical safety and regulatory duties attached. We keep the assistant to administrative work, and your advisers confirm where it sits.
Retrieval-augmented generation
A way of building an AI assistant so that it looks up the relevant passage in your own approved material before answering, rather than relying on what the model learned elsewhere. It is what lets an assistant give your parking directions in your own words, and say it does not know when nothing in your material covers the question.
Human handover
The point where the assistant stops and passes the conversation to a person, with everything said so far attached. In a clinic it is the most important feature, not a fallback: anything clinical, personal or unclear should reach reception quickly, with the patient told when to expect a reply.

Handled with care

The details that matter here

An assistant speaks for your practice when nobody is watching. These are the details we settle with your clinical lead and data protection lead before it speaks to anyone.

  • The boundary is tested, not assumed

    With your clinical lead we write a set of test messages: symptoms, medication questions, results, distress, someone who may be at risk, and the indirect ways people describe them. The assistant must stop and hand over on every one. The tests run before launch and after every change, and your clinical lead signs off the results.

  • Urgent help wording owned by clinicians

    What the assistant says when it stops, including where to get urgent help, is written and approved by your clinical lead, not by us and not generated by the model. It appears word for word, every time, and changes only when your clinical lead changes it.

  • No personal details without verification

    Confirming that someone is a patient is itself personal information. We design the assistant to reveal nothing about a person or their appointments until identity is verified, and agree with your data protection lead how parents, carers and people booking for someone else are handled.

  • Where conversations go

    We agree with your data protection lead which model provider is used, where, and for how long conversations are kept. Health data is special category data under UK and EU GDPR, and in the US a supplier handling protected health information works under a business associate agreement. Conversations are not used for training.

Guides

Worth reading first

Questions

What buyers ask first

Can an AI receptionist give medical advice?

Ours does not, by design. It is built and tested to stop as soon as a message mentions symptoms, medication, results or anything about someone's care. It shows the urgent help wording your clinical lead has written and hands the conversation to reception, who follow your own procedure. It does not try to judge how serious the question is. Every such message is treated as one for a person.

Can an AI receptionist access patient records?

Ours cannot. It can read the diary to find free slots and, for a verified patient, book or move their own appointments. It cannot see clinical notes, letters, results or invoices, and it cannot tell anyone what is in them. Giving it less access is deliberate: an assistant that cannot reach clinical information cannot repeat it to the wrong person.

Can an AI receptionist answer the phone for a clinic?

It can, but we would not start there for a clinic. Callers are often worried or unwell, and their first sentence is often about how they feel, which is exactly what the assistant must hand over. If you want a voice line later, we would limit it to opening hours, directions and booking for verified patients, and pass every other call to a person.

What happens when an AI receptionist gets something wrong?

It can, so we design for that. It answers only from the library your team approves, and when nothing there fits it says so rather than guessing. Every conversation is logged and reviewed, wrong answers are traced to the library entry that caused them, and the fix is approved by your team. It starts by drafting replies that reception approves before sending, until you are comfortable letting it answer directly.

Do patients have to be told they are talking to an AI?

We recommend it everywhere, and in some places it is required. In the EU, the AI Act's transparency rules, which apply from 2 August 2026, require people to be informed when they are interacting with an AI system unless that is obvious. Our assistant says it is automated at the start of every conversation and offers a person at any point, with wording agreed with your compliance lead for each region you serve.

Send us the questions reception answers most

List the questions your front desk hears every day. We reply within 48 hours and will tell you which ones an assistant should answer and which must always go to a person.