# Is an AI receptionist worth it? The questions to ask first Picture an office manager at a dental practice with three vendor websites open side by side. Each one says it answers, books and hands over to the team, in nearly the same words. Short answer: an AI receptionist is worth it when your clinic loses bookings to calls nobody can pick up, and the one you choose can show you what it did. The questions below help you check both, whichever vendor you are talking to. ## Is an AI receptionist worth it for your clinic? Two things decide it. The first is how much your clinic loses to calls and messages nobody answers in time. The second is whether the system does the job well enough that your team trusts it with patients. The first you can count. The [cost of missed calls](https://agentblue.io/guides/cost-of-missed-calls) guide shows how to work it out from your call log, and the [missed call calculator](https://agentblue.io/tools/missed-call-calculator) turns your own figures into a range. If the range is small, you can stop reading here with a clear conscience. The second you have to check, and that is what the questions are for. Nobody hands a house key to a dog sitter because their website looked nice. You ask who comes round, what happens if the dog gets out, and who you call if it does. Buying an AI receptionist is the same conversation with a clinic attached. The questions come in five groups. Ask every vendor all of them, in the same order, and write down what each one says. ## What will it answer, and what will it leave alone? Start with the job itself. Vague answers here usually mean the vendor has not thought much about your kind of clinic. - Which channels does it answer: phone calls, texts, web chat, or some of these? - Which hours does it cover, and can we change them ourselves? - Does it book straight into our booking system, or does it send requests for the team to confirm? - Which of our services can it book, and which ones always go to a person? - What does it do with a question it was not set up to answer? - Does it tell callers it is an AI? - How does it handle a caller who wants two things at once? What a good answer sounds like: specific, and about your clinic. “It books cleanings and new-patient exams directly, sends implant consults to your coordinator, and tells callers it will check with the team on anything about treatment.” You can check an answer like that on a test call, which is the point of asking. ## How does it hand a caller to your team? - How does a caller reach a person, and how soon is that offered? - What does our team see when it hands over: the caller’s details, the reason for the call, what has been said so far? - What happens after hours, when there is nobody to hand to? - Who hears about an urgent call, and how? - Can we set our own rules for which calls always go to a person? Ask to see a handoff on the demo, from the caller’s side and from your team’s side. The caller should not have to repeat themselves, and the note your team receives should make sense to someone who was not on the call. [Warm transfer](https://agentblue.io/glossary/warm-transfer) explains the live version, where the call itself is passed across. The after-hours answer deserves a follow-up question. If the system promises callers a callback, ask who sees that promise in the morning and where. ## Who sets it up, and who keeps it right? - Do you set it up for us, or do we set it up ourselves? - What do you need from us, and how much of our team’s time does it take? - How long from signing to live, and what starts that clock? - Who updates it when our hours, services or team change? - Who do we contact when it gets something wrong, and how soon is it fixed? - How do we test it before it talks to real patients? A self-serve tool leaves the setup and the upkeep with your team. A done-for-you service puts both on the vendor. Either can work, as long as you know which one you are buying and who will update it the week you add a new treatment. The last question on the list matters most. A vendor that welcomes your test calls is telling you something. The guide on [how to test an AI receptionist](https://agentblue.io/guides/how-to-test-an-ai-receptionist) has the scripts and a scoring sheet. ## What exactly are you signing up for? - How is it priced: per call, per minute, a flat monthly fee, or a managed service? What makes the bill go up? - Is setup included, and what about changes later? - How long is the minimum term, and how much notice do we give to leave? - Is there a trial or a pilot period, and what happens at the end of it? - If we leave, what happens to our phone number, our call records and the patient details it holds? - Which extras cost more: more channels, more locations, longer hours? Ask for these answers in writing and put them side by side. Pricing models differ between vendors, so two quotes that look alike can add up differently by the end of a busy month. ## How does it show what it knows about a patient? This is the group that is easiest to skip, and the one that tells you the most. An AI receptionist that greets a returning patient by name sounds impressive. The real questions are what it knows about that patient, where it learned it, and what it does when it is not sure. - When it uses a fact about a patient, can it show where that fact came from: the booking system, an earlier call or text, a note from our team, or the patient’s own words? - Can it show when it learned each fact? - Can our staff see everything it holds about a patient, and correct or delete any of it? - When two sources disagree, for example the booking system has one phone number and the patient gives another, which one does it use, and does it ask? - What does it do when it does not know something? Can you show me it saying so? - How does it keep two people apart when they call from one family phone? - Does it write anything back into our booking system or practice software, and can we see what it wrote? - How do you check that it is not making things up, and can we see how you check? What a good answer sounds like: the vendor opens a sample patient in the demo and points at each fact, its source and its date. A weak answer is a promise that it simply remembers, with nothing to point at. Why push on this? An AI that answers from a wrong fact sounds exactly as sure of itself as one that answers from a right fact. The only way your team can tell the difference is to see where each fact came from. A shop receipt is useful for the same reason: you can check it against what is in the bag. To hear the difference on a real call, the returning-patient calls in [how to test an AI receptionist](https://agentblue.io/guides/how-to-test-an-ai-receptionist) put these questions to work. ## How should you use these questions on a demo? Send the list to every vendor before the demo, so the call is spent on answers rather than slides. Then split it: some answers you want to see, and some you want in writing. Which answers to see on the demo and which to get in writing | Question group | Ask to see it on the demo | Ask for it in writing | | --- | --- | --- | | What it answers | A test call about a service it should pass to a person | The channels, hours and services it covers | | Handoff | A handoff from both sides, the caller’s and your team’s | Who hears about urgent calls, and how | | Setup | Where your clinic’s facts are kept and how they are changed | Who does each part of the setup, and how long it takes | | Contract | Nothing to see; this group is paperwork | Pricing model, minimum term, notice, and what happens to your records if you leave | | What it knows about a patient | A sample patient with each fact, its source and its date, and a staff member correcting one | How it handles details that disagree and calls from a shared phone | Score the answers the way you would score test calls: the same questions, in the same order, for every vendor, with notes written straight after the call. A demo where the vendor kept steering back to the slides is an answer too. ## When is an AI receptionist not worth it? The honest catch: for some clinics the answer is no, or not yet. It usually looks like one of these: - One person handles your phone comfortably, and your call log shows few missed calls. - Most of your calls are clinical questions that need a clinician, so there is little a system should answer on its own. - Your booking system cannot be connected, and your team would end up typing every booking in again by hand. - You want it to replace people entirely. Some calls need a person, and a system that never hands over will upset the patients who make them. - You have no time to test it properly before it goes live. If a couple of these sound like your clinic, spend the money elsewhere for now and recount your missed calls in a few months. Clinics change, and so does the count. ## How does AgentBlue answer the questions about what it knows? Since we are asking you to put these questions to every vendor, here are our answers to the group about what it knows. We call it memory with receipts. - When our AI uses a fact about a patient, it shows where that fact came from and when. - Your staff can see each fact, correct it and delete it. - When it does not know something, it says so instead of guessing. - It keeps each patient’s history with your clinic, so a returning patient does not start from scratch. - It tells patients it is an AI, and hands a patient to your team with the details so far. The rest of the list is for the demo call. Our [AI receptionist](https://agentblue.io/solutions/ai-receptionist) is done for you: we set up, connect and run it, and it is live in about 2 weeks after we receive access to your systems. Pricing is tailored to your clinic and shared on the free demo call. Bring this list and ask every question on it. ## What should you do this week? - Count last week’s missed calls, or put a tally sheet by the phone. - Copy the questions on this page into one document and delete the ones that do not apply to your clinic. - Send the same document to each vendor on your shortlist before their demo. - Plan test calls for the week after, with the same scripts for everyone. Once the answers are in, ask yourself one question: which vendor showed you something, and which one only told you? ## Frequently asked questions ### Are AI receptionists good? Some are good at some jobs. The useful question is whether one is good at your clinic’s calls, and a short round of test calls answers that better than any review. ### Does an AI receptionist work for a small clinic? It can, when the clinic loses bookings to calls nobody can pick up, often at lunch, in busy treatment blocks and after closing. If one person comfortably handles your phone and your missed call count is low, it may not be worth it yet. ### What is the first question to ask an AI receptionist vendor? Ask what it does when it does not know the answer, and ask to see it happen. The reply shows how the system was built, and how much of the demo you can take at face value. ### Will an AI receptionist replace my receptionist? Plan for it to take the calls your team cannot get to, and check that it hands the others to a person with the details so far. Your team keeps the conversations that need a human, and gets back some of the time the phone used to take. ### Why ask an AI receptionist where its facts came from? Because a wrong fact said with confidence sounds the same as a right one. If the system can show the source and date of each fact, your team can check it and fix it. If it cannot, every answer is taken on trust. ## Related resources - [Trust](https://agentblue.io/trust) ## Book your free demo [Book a demo](https://agentblue.io/demo) ### What to expect - Free, 30 minutes, no obligation - Pricing tailored to your clinic - A person replies within one business day, Monday to Friday - Limited onboarding slots each month Source page: