Guide
How do AI receptionists work? A plain-English explainer
An AI phone receptionist listens, works out what the caller wants, answers from your own information and books or routes the request. This is what happens in each of those steps, and what you need to set up.
Desk · AttractKey takeaways
- An AI receptionist turns speech into text, works out the request, answers from your information, takes an action and speaks the reply.
- It is only as good as the information you give it: hours, services, prices, policies and a live calendar per location.
- The handover rule (what goes to a person) is a business decision you make before launch.
- The same engine can answer calls, WhatsApp, website chat, forms and email.
- In the US, outbound calls with an AI voice fall under the TCPA; inbound answering is the common starting point.
How does an AI receptionist work, in short?
An AI receptionist answers your phone (and often your chat and messages), converts what the caller says into text, uses a language model to work out the request, answers it from your own business information, and speaks the reply in a natural voice. When the request needs an action, it books the appointment, takes the message or transfers the call to a person by the rules you set.
That is the whole loop, repeated every turn of the conversation, usually within a second or two. The rest of this guide opens each step. For the one-line definition, see what is an AI receptionist or the glossary entry on the AI receptionist.
- 01Caller speaksPhone, in their language
- 02Speech to textWords transcribed live
- 03UnderstandWhat do they want?
- 04Look upYour hours, prices, calendar
- 05ActBook, answer, take a message
- 06SpeakReply in a natural voice
- 07LogTranscript and summary saved
This loop runs on every turn of the call, not once per call.
The parts inside an AI phone receptionist
Most voice AI receptionist products are built from the same five parts, whether they are sold to a dentist or a car dealer. Knowing the parts helps you ask better questions in a demo.
| Part | What it does | What to ask the vendor |
|---|---|---|
| Telephony | Receives the call on your number (forwarded or ported) and can transfer it. | Do I keep my number? Can it transfer to a person and pass the context? |
| Speech to text | Turns the caller's voice into text as they speak. | Which languages and accents? How does it handle a noisy line? |
| Language model | Works out intent, asks follow-up questions, writes the reply. | What stops it from answering things it does not know? |
| Knowledge and tools | Your information (hours, services, policies) and connections to a calendar, CRM or booking system. | Which calendars and CRMs connect? Per location? |
| Text to speech | Speaks the reply in a natural voice. | Can I pick the voice and the greeting? |
On top of these sit the business rules: what to answer, what to collect, when to hand over, and where to send the summary. These rules are where most of your setup time goes, and they matter more than the model.
Where the answers come from
A well-built AI receptionist does not make answers up from general knowledge. It answers from the information you give it, and says it does not know when the information is missing. That information is per location, because hours, staff, services and prices often differ by branch.
Information an AI receptionist needs per location
This is the same set of facts your Google Business Profile and listings carry. If your listing hours are wrong, the AI receptionist will be wrong too. Run the free Google Business Profile audit to check the basics before you load them.
The free audit scores your Google Business Profile, flags missing or inconsistent details and shows your Maps rank against the top three.
What an AI receptionist can actually do on a call
Answering questions is the easy part. The value is in the actions it can take without a person: these depend on what you connect.
Common actions
Checks the connected calendar, offers times, confirms in writing by text or email.
Hours, directions, services, prices you publish, policies.
Asks the questions you set (service, location, budget, timing) and logs the answers.
Records the caller, reason and callback time, and sends a summary to the right person.
Hands the call to a person by your rule, ideally with a short summary first (a "warm" transfer).
Sends a confirmation, reminder or missed call text back after the call.
- Can it be answered from your information?NoSay so, collect details, hand over to a person.Yes
- Is it a topic you allow it to handle (not a complaint, refund, medical or legal question)?NoHand over to a person with the transcript.Yes
- Does it need an action (book, change, cancel)?NoAnswer it and offer the next step.YesDo the action, then confirm it in writing.
You decide the topics that always go to a person.
For the list of topics most businesses keep with people, read what an AI receptionist should never answer on its own.
Is it only for phone calls?
No. The same engine (understand, look up, act, reply) works on text channels too, without the speech steps. Many businesses get more value from one assistant answering every channel than from a phone-only tool, because the information and the inbox stay in one place.
- Calls. Voice, with transfer to a person. See the AI answering service.
- WhatsApp and SMS. Text replies, booking links, reminders. See WhatsApp.
- Website chat. Answers visitors on the page they are reading. See website chat and the roundup of the best website chat for small business.
- Forms and email. Replies to an enquiry in minutes instead of the next morning. See forms.
AI receptionist use cases by industry
The engine is the same everywhere; the questions and the handover rules change. Here is how it is commonly set up across industries.
| Industry | Handles on its own | Always hands to a person |
|---|---|---|
| Dental and clinics | New patient booking, hours, insurance accepted, directions | Symptoms, clinical questions, billing disputes |
| Home services | Service area check, booking a visit, emergency triage questions | Quotes on complex jobs, complaints |
| Auto dealers and service | Service booking, test drive requests, hours by department | Trade-in values, finance terms |
| Restaurants | Reservations, hours, menu questions, large-party enquiries | Allergies beyond the published menu, complaints |
| Hotels | Availability from the booking system, check-in times, parking | Refunds, special arrangements, group contracts |
| Real estate and property | Showing requests, listing basics, office hours | Offers, negotiations, legal questions |
| Retail | Store hours, stock questions if connected, order status | Returns outside policy, complaints |
Hotels have a few extra steps; see can an AI receptionist take a hotel booking. If you run several sites, the pillar on the AI receptionist for multi-location businesses covers rollout by location.
In healthcare, legal and financial services, decide with your compliance team what an AI receptionist may say, collect and store. Ask vendors about data handling and, where relevant, whether they sign a business associate agreement.
What AI receptionists still get wrong, and the rules that apply
They are good at routine requests and weak at anything that needs judgment, empathy or information they were not given. The failures you hear about usually trace back to missing information or a vague handover rule, not to the voice.
- Missing information. If the holiday hours are not loaded, it cannot know them. Update them per location.
- Unclear audio. Bad lines and heavy background noise cause misheard names and numbers. Good systems read numbers back to confirm.
- Over-confidence. Settings should make it say "I do not know, let me get someone" rather than guess.
- Disclosure. Decide whether your AI says it is an AI at the start of the call. Many businesses choose to, and some vendors say theirs never claims to be human when asked.
On rules: the FCC has ruled that calls using AI-generated voices count as "artificial" voice calls under the TCPA (FCC). That matters most for outbound calls, such as reminders or follow-ups, which need the consent the TCPA requires. Recording and transcribing calls is also covered by state laws, and some states require every party to agree. Tell callers the call may be recorded, and check your setup with your counsel.
Mistakes to avoid
- Launching with only the website copy loaded
Instead: Load per-location hours, services, prices you share, policies and a live calendar. Gaps become wrong answers.
- No handover rule
Instead: Write down the topics that always go to a person, and who receives them, before launch.
- Judging it on a scripted demo
Instead: Call the test line with odd requests, interruptions and a request for a person.
- Using it for outbound calls without checking consent
Instead: Start with inbound answering. Ask your counsel before any outbound AI-voice calls.
- Not reading the transcripts
Instead: Read a sample every week for the first month and fix the information it lacked.
How to set one up: the first 30 days
- Day 1 to 3
Pull a month of calls and messages, sort them by request type, and pick the top five to automate.
- Day 4 to 7
Load per-location information, connect the calendar, write the handover rule and greeting.
- Week 2
Run it on after-hours calls only. Read every transcript.
- Week 3
Fix the gaps, then add busy-hour overflow and one text channel.
- Week 4
Review bookings, handovers and missed calls per location; extend to more locations.
Always hand over to a person when the caller:
- has a complaint or asks for a refund
- describes a medical, legal, safety or financial concern
- asks for a person by name or asks for a manager
- asks something not covered by our information
During business hours: transfer to {role} at {location}, with a one-line summary.
Outside business hours: take the name, number and reason, tell the caller when we will call back ({time}), and send the summary to {inbox}.Write one per location if the people and hours differ.
See an AI receptionist work on your own informationDesk by Locus Intelligence answers calls, WhatsApp, website chat, forms and email for every location from each location's own information, books appointments, and hands over to a person by your rule.
Comparing tools? Read the best AI receptionist buyer's guide. Not ready for a full receptionist? Start with missed call text back.
Google Business Profile auditFree. You get a profile score, your Maps rank against the top 3, and the fixes that matter most.
Your next steps
Today
- Export last month's calls and messages and count them by request type and hour.
- Run the free Google Business Profile audit to check the hours and details your receptionist will use.
This week
- Write your handover rule using the template above.
- List per-location hours, services, prices you share and policies in one document.
- Book two or three demos and call each test line with hard requests.
This month
- Pilot on after-hours calls at one or two locations.
- Read transcripts weekly and fix missing information.
- Add a text channel such as WhatsApp or website chat.
Frequently asked questions
How do AI receptionists work?
They transcribe the caller's speech, use a language model to understand the request, answer from your business information, take an action such as booking, and speak the reply. The loop repeats every turn.
Can callers tell it is an AI?
Modern voices sound natural, but many callers can tell, and many businesses choose to say so at the start. Decide your disclosure policy before launch.
Can an AI receptionist book appointments?
Yes, when it is connected to your calendar or booking system. Without that, it takes the request and routes it to your team.
What happens when it does not know the answer?
A well-configured one says so, collects the details and hands over to a person with the transcript.
Is an AI receptionist the same as an answering service?
Not quite. An answering service usually takes messages; an AI receptionist resolves routine requests. See AI receptionist vs answering service.
Is a virtual receptionist an AI?
Not always. A virtual receptionist can be a remote human team or software.
Related
All guidesSoftware that answers calls and messages from a business's own information, books or routes the request, and hands over to a person when needed.
Glossary · Desk · 2 minWhat is an AI receptionist?An AI receptionist is software that answers a business's calls and messages when nobody else can, understands what...
Question · Desk · 2 minWhat should an AI receptionist never answer on its own?Anything that needs judgment or carries risk: complaints, refunds, medical or legal advice, safety issues and...
Question · Desk · 2 minVirtual receptionistA person or service that answers calls remotely for a business. An AI receptionist is the software version.
Glossary · Desk · 2 minMissed call text back: setup and scriptsHow it works, five scripts to copy, and the US texting rules to check.
Guide · Desk · 10 minNever miss another customer.
Start with a free audit. See what each location is losing, and where.
Profile score, Maps rank against the top 3, and the fixes that matter most.
Or book a demo- Role-based access by region and brand
- Every action logged and reversible
- Nothing customer-facing goes live without approval