AI receptionist, answering service or text-back: which one fits a small business.

The short answer

Three things can catch the calls you miss. An answering service puts a trained person on the line in your business name. An AI receptionist answers with a computer voice that handles routine questions and some bookings. Missed-call text-back lets the call ring out as it does now, then texts the caller straight away from your number.

Choose by what your callers need in the minute after you cannot answer:

• They want a booking, a quote or a call back, and can wait a little: text-back, at the lowest cost.
• They need a human voice now, such as a worried patient, a tenant with a flood or a new legal client: a live answering service.
• You get many routine calls at the same moment, such as a restaurant on a Friday evening: an AI receptionist, tested hard and introduced honestly.

GGSites sells missed-call text-back, not a voice agent. Read this guide with that in mind.

Count your missed calls first

For two weeks, note every call you could not answer: the time, whether they left a voicemail and what they wanted when you rang back. Your phone system's call log gives you the first two. The third comes from the call back.

Then sort them:

• Routine: opening hours, prices, directions, a booking to move.
• New business: a quote, a first appointment, a new client.
• Urgent: something that cannot wait an hour.
• Noise: sales calls and wrong numbers.

The mix decides the answer. A plumber whose missed calls are mostly new jobs, with the odd emergency, needs something different from a dental practice whose missed calls are mostly patients moving appointments. Without this list you are choosing on the strength of a sales demo.

Fix the basics first

Whatever you add, these cost nothing and help every option work better:

• A voicemail greeting that gives your name, says when you ring back and gives the emergency route, if there is one.
• One person each day who owns call backs, with a named backup.
• A call-back promise you can keep on your busiest day, such as "by the end of the day".
• Your opening hours and booking link on your website and Google Business Profile, so some callers never need to ring.
• A written list of what counts as urgent in your business, so whoever or whatever answers knows when to interrupt you.

A live answering service

Operators at a call centre answer in your business name, working from a script you agree. Depending on the service, they take a message, book into your calendar, answer a set list of questions or put urgent calls through to your mobile.

Where it wins: a person can calm an upset caller, hear that a "quick question" is really an emergency, and cope with someone who struggles with machines. Where it falls short: the operator knows your business only as well as the script, messages vary with whoever took them, the bill grows with every call or minute, and at busy times callers can wait on their side too.

Questions to ask before you sign:

• Who answers, and do they work from my script or a general one?
• How do urgent calls reach me, and what happens if I do not pick up?
• Can they book straight into my calendar or booking software?
• How do messages reach me: email, text, an app?
• How is the bill counted: per minute, per call, with a monthly minimum?
• Do you record calls, and who keeps the recordings?

An AI receptionist

An AI receptionist answers every call at once with a computer voice. It can answer questions from information you approve, take a message, book where your booking software allows and transfer the call to a person. It does not get tired, it does not put people on hold, and it can take several calls at the same time.

Where it struggles: background noise, strong accents and bad lines make it mishear names, numbers and addresses. If its information is thin or out of date, it can give a wrong answer in a confident voice. Some callers hang up as soon as they hear a machine, and others ask for a person straight away, so the route to a human has to be one sentence away.

It needs the same work as a website chatbot: the questions your callers ask, approved answers, a list of what it must never say, handover routes, and testing with real calls. What a chatbot should hand over applies to a voice on the phone as much as to a chat window.

Questions to ask an AI receptionist provider:

• What does it answer from, and how do I correct an answer?
• How does it hand a call to a person, and what happens if nobody picks up?
• Can it book into my booking software, or only take a request?
• Does it say it is an AI, and can I change the greeting?
• Does it record or transcribe calls, where are they kept and for how long?
• What does a call cost once the plan's minutes run out?

If it also rings people, such as calling back missed callers or ringing with reminders, the FCC has confirmed that an AI-generated voice counts as an artificial voice under the TCPA. Those calls need the person's consent first.

Missed-call text-back

Calls ring exactly as they do now. When one goes unanswered, the caller gets a text from your business number straight away: who you are, that you will call back, and a way to reply or book. The reply lands where your team already reads messages.

Where it wins: it costs the least of the three, nothing pretends to be you, and the caller gets your name and a booking link in writing. For a caller who wanted a booking or a quote, a fast and clear text is often enough.

Where it falls short: it does not answer the call. A caller on a landline cannot receive the text. Someone still has to answer the replies, and an emergency still needs a person. Your number has to take texts and be registered with the US carriers, and the consent rules for texting apply. Texting from your business landline covers the number side.

GGSites sets this up as part of its automation work, with the US carrier registration handled.

What callers will put up with

There is no single answer, so take your own callers one at a time:

• A regular moving a booking: fine with any of the three, and happiest with a link to do it themselves.
• A new customer ringing three plumbers: books with whoever responds first and clearly. A text that arrives at once can do that, and so can a voice that takes the details properly.
• An older patient: often wants a person. A machine that asks for a date of birth twice loses them.
• Someone with a burst pipe or a tenant locked out: needs a person or a direct line to the on-call phone, whatever else you use.
• A caller on a building site or driving: hears little and says less. A text they can read later works better than a voice asking questions.
• A new legal or financial client: expects discretion and a human. Use a live service, or a text that books a call, not a voice that asks about their case.

What decides the cost

Prices vary too much between providers to quote, so compare what each bill is made of:

• Answering service: per minute or per call, sometimes with a monthly minimum, and extra for out-of-hours cover, booking or putting calls through.
• AI receptionist: a monthly plan with included minutes or calls, a charge for each extra minute, and the setup time for its information, booking connection and testing.
• Text-back: a setup, a monthly plan, carrier registration and a small cost per text.

Then add the cost in your own time: reading messages or transcripts, fixing wrong bookings and ringing people back. A cheap service that hands you garbled messages costs more than it saves. Our own setup and monthly prices for text-back are on the prices page.

Recording calls, and saying it is an AI

Answering services often record calls, and an AI receptionist usually works from a recording or a transcript of each one, since that is how you check its work. Ask whether yours records, where the recordings are kept and for how long.

US federal law allows a call to be recorded when one person on it agrees, but a number of states, including California, Florida and Pennsylvania, require everyone on the call to agree. Callers ring from anywhere, so the simple habit is to say it at the start: "This call may be recorded." In Europe the GDPR covers recordings and transcripts: tell callers, have a reason to keep them and delete them on a schedule.

Say plainly that the voice is an AI, too. Under the EU AI Act, from 2 August 2026 people must be told when they are talking to an AI system unless it is obvious, and some US states have disclosure rules of their own. A greeting such as "You are through to the automated assistant for Harbour Dental" is honest and costs nothing.

This is general information, not legal advice. If you record calls or use an AI voice, ask an adviser which rules apply where your callers are.

Using two together

They combine well, and a mix often beats any one of them:

• Text-back during the day, when calls are routine and you ring back within the hour, and an answering service at night for emergencies only.
• An AI receptionist for questions and bookings, with any call it cannot finish put through to a person, and text-back as the fallback when nobody picks up.
• Text-back for everyone, plus a separate emergency number given to existing customers, such as tenants or patients after a procedure.

Test it before your callers do

Whatever you choose, ring it yourself, from different phones, with these calls:

• A routine question, such as parking or opening hours.
• A price question, which it should answer from your list or decline.
• An emergency, said calmly and then in a panic.
• A name and address with an unusual spelling, from a noisy street.
• "Are you a real person?"
• A caller who wants to complain.
• A call from a landline, to see what text-back does with it.
• A call one minute before you close, and one a minute after.

Then read what reached you. The message, transcript or text should let you ring back without asking a single question again.

Four businesses, four choices

• A two-van plumbing firm whose missed calls are mostly new jobs: text-back for every missed call, and an on-call number for emergencies in the voicemail greeting. The office books visits from the replies.
• A dental clinic with one receptionist: text-back for missed calls in clinic hours, with a link to book or move an appointment, and an answering service after hours that passes dental emergencies to the dentist on call.
• A small law firm: a live answering service that takes a new client's name, number and type of matter and books a call, without asking about the case. Text-back covers existing clients.
• A busy restaurant: an AI receptionist for opening hours, directions and table requests on Friday evenings, putting large parties and complaints through to the manager.

Mistakes to avoid

• An AI voice that pretends to be a person, or dodges the question when asked.
• Letting any of them quote prices or promise times you have not approved.
• No route to a person, or a route that ends in voicemail.
• Sending every call to the service, including the ones you could have answered.
• A long greeting full of options before the caller can speak.
• Not reading the messages, transcripts or replies after the first week.
• Forgetting that landline callers never see a text-back.

If the website gets the same questions as the phone, a chatbot may take some of the load first. Chatbot, live chat or contact form compares the options for the website side.