AI Receptionist for Small Business: What It Does and What It Costs

You are cutting someone's hair, fixing a boiler, or with a patient when the phone rings. It goes to voicemail, and the caller rings the next business on the list instead of leaving a message. Website visitors do the same: a question at 9 p.m. with no answer becomes a competitor's enquiry. A receptionist catches those people, and most small businesses cannot justify a full-time one.

This guide is for owners of small service businesses, such as salons, clinics, trades, agencies and law offices, who are weighing up an AI receptionist. It explains the three things sold under that name, what they cost on the vendors' own pricing pages as of October 2026, how to set one up, and the risks to plan for. It ends with where PepoChat fits, including what its voice option does not do yet.

What is an AI receptionist?

An AI receptionist is software that does a receptionist's front-desk work on your phone line or website: it answers, helps with simple requests and routes the rest. The term covers three different products, and the difference matters more than any feature list.

An AI receptionist is an automated agent that answers incoming calls or chats in your business's name, using a language model and your own information to answer questions, take messages, book appointments and transfer people to your team. It is not a phone tree. A phone tree makes callers press 1 or 2; an AI receptionist understands "I need to move my Thursday appointment" in the caller's own words.

The three things you will find under the name are:

• An AI phone receptionist, a voice agent that answers your business number. It listens, transcribes, decides and speaks back with a synthetic voice. Vendors such as Rosie, Goodcall and Smith.ai's AI Receptionist sell this.
• A website chat agent, an AI agent in a chat window on your site. It does the same front-desk jobs in text, for visitors who are already on your pages. PepoChat is in this group.
• A virtual receptionist service, a team of real people at a company such as Ruby or Smith.ai who answer your calls or chats remotely from a script you write. Many now use AI for transcripts and call notes, but a human does the talking.

Most small businesses need one of the first two plus a clear route to a person, chosen by the channel their customers actually use.

What does an AI receptionist actually do?

An AI receptionist does five jobs: it answers frequent questions, takes messages, qualifies enquiries, books appointments, and routes people to the right person. Everything else, such as refunds, complaints and judgment calls, should reach a human.

Answer frequent questions

Most calls to a small business are the same dozen questions: opening hours, prices, location and parking, whether you do a particular service, how long it takes, what to bring. An AI receptionist answers these from the information you give it, so a thin website means a lot of "I'm not sure", and a bare prompt invites guessing. Insist on a product that answers from your own content and admits when it does not know.

Take messages

When a caller needs you personally, the receptionist takes a message: name, number or email, what it is about, and how urgent it is. Then it gets that message to you by email, text or a notification. Answering machines fail at this because many callers never leave a voicemail; a receptionist that asks the questions gets a usable message instead of a beep.

Qualify enquiries

Qualifying is asking the few questions that tell you whether an enquiry is worth a call back and who should take it: which service, where the job is, roughly when, sometimes a budget. A plumber wants to know the postcode and whether water is leaking now. A law office wants to know the type of matter before anyone books a consultation. Keep the list short. Three questions feel like help; eight feel like a form.

Book appointments

A good AI receptionist checks live availability on your calendar, offers a few times, and books the one the customer picks, with a confirmation email or text. The details that go wrong are time zones, double booking and missing contact details, which the guide to AI chatbot appointment booking covers step by step. If a product "takes the request" for you to confirm later, it is a message taker, not a booker.

Route to a person

The last job is knowing when to stop. A receptionist should pass the caller or chatter to a person when they ask for one, when it cannot answer, when the customer is upset, and for anything involving money or exceptions. On a phone line that means a live transfer or a promised call back; in chat it means a handoff to your inbox with the conversation attached. The human handoff guide explains what a good escalation looks like.

How much does an AI receptionist cost in 2026?

As of October 2026, AI phone receptionists cost roughly $49 to $150 a month for small volumes, human virtual receptionist services start around $250 to $300 a month, and website chat agents range from free tiers to flat plans of about $49 a month. The billing unit differs by vendor, which makes headline prices hard to compare.

These are list prices from each vendor's own pricing page, checked in October 2026:

The sources: Rosie's pricing page also lists a $299 Growth plan with 2,000 minutes and a 7-day trial. Goodcall's pricing states it has no per-minute charges and counts a unique customer as one phone number, once per month. Smith.ai's AI Receptionist pricing lists larger bundles at lower per-call rates, such as $1.80 a call at 150 calls, and says calls from known spammers are filtered out before they count. Its human receptionist plans run from $300 for 30 calls to $810 for 90 calls, with 24/7 answering at the same price day or night. Ruby's plans go from $250 for 50 minutes to $1,725 for 500 minutes, with website chat from $143 for 10 chats.

Per call, per minute, per caller or per reply

The billing unit decides which product is cheap for you. Per-minute plans (Rosie, Ruby) favour short calls. Per-call plans (Smith.ai) favour long ones. Per-caller plans (Goodcall) favour businesses whose regulars call several times a month. Per-reply plans, which is how PepoChat counts, charge for each AI answer in a chat, so a three-message exchange costs three replies. Before you compare, pull one month of your call log and note three numbers: calls, total minutes and distinct numbers.

A worked example: 150 calls a month

Take a salon that gets 150 calls a month, averaging three minutes, from about 100 different numbers. That is 450 minutes. At October 2026 list prices:

At this volume human answering costs roughly ten times as much, and buys a person who can calm an angry caller and use judgment. Whether that is worth $1,000 or more a month depends on what one lost call costs you. A missed call for a $40 haircut and a missed call for a $4,000 kitchen job are different decisions.

Add the costs that are not on the price card: your hours writing the receptionist's information and reviewing early calls, plus any calendar or phone-number add-ons. On developer voice platforms a minute is billed by several providers at once; the AI voice agents guide breaks that down.

How do you set up an AI receptionist?

You can set up an AI receptionist in an afternoon: list what people ask, write down the answers, decide what it may do on its own, connect your calendar, choose how it reaches you, then test it as a customer would. The writing is most of the work.

1. List what people actually ask

Spend a week noting every call and chat: what they wanted and how it ended. Your call log, your inbox and your own memory are enough. You will find that a dozen questions make up most of the volume. Those are the receptionist's first job.

2. Write the answers down

Put opening hours, services, prices or price ranges, locations, parking, cancellation rules and the answers to your dozen questions on your website or in a document. Website chat agents can read your site directly; phone receptionists usually take a document or a form. Be specific. "Prices vary" produces a receptionist that says "prices vary".

3. Decide what it may do alone

Write three short lists: what the receptionist may answer, what it may do (book, take a message, send a link), and what it must always pass to you (refunds, complaints, medical or legal questions, anything urgent). Include the things it must never say, such as never promising a price that is not on your list.

4. Connect your calendar

Booking is where an AI receptionist saves the most back-and-forth. Use a scheduling tool that checks live availability and re-checks the slot when the booking is made, so two callers cannot take the same time. Offer one or two appointment types first, not your whole menu.

5. Choose how it reaches you

Decide where messages and handoffs go: email, text, a team chat, or a shared inbox. Decide what the receptionist promises outside your hours. "Someone will call you back tomorrow by 10 a.m." is a promise a customer can plan around; "shortly" at 11 p.m. is not. The guide to 24/7 support without a night shift covers business hours and away messages in detail.

6. Test it as a customer

Call or chat with it yourself, then ask a friend to. Ask the dozen questions, then the awkward ones: a request you do not offer, an angry complaint, a question with no answer in your content, a booking on a day you are closed. Fix the gaps in your information, not in the receptionist's personality. For the first month, read or listen to a sample of conversations each week; every question it could not answer is a line to add.

What are the risks and limits of an AI receptionist?

The main risks are wrong answers, customers who feel misled about talking to a machine, recording calls without the consent your state requires, and urgent requests that sit in a queue. Each has a simple control, set up before you go live.

Wrong answers. A language model can sound confident while guessing. Choose a product that answers only from your information and says "I don't know" otherwise, keep that information current, and put prices and policies in writing rather than in the prompt.

Disclosure. Tell people they are talking to an AI. It is increasingly a legal duty: the EU AI Act's transparency rules have applied since 2 August 2026, and several US states have their own rules. The AI disclosure laws guide summarises them. It is also good practice: callers who discover halfway through that they have been talking to a bot feel tricked, and a one-line introduction such as "Hi, I'm the AI assistant for Bright Smile Dental" costs nothing.

Call recording. If your phone receptionist records calls or keeps transcripts, check the consent rules where you and your callers are. The Digital Media Law Project's guide to recording phone calls explains that federal law in the US is "one-party consent", while a group of states, including California, Florida, Pennsylvania and Washington, require the consent of everyone on the call. For calls across state lines it advises getting everyone's consent. The simplest control is a greeting that says the call may be recorded.

Urgent requests. A receptionist that takes a message at 2 a.m. about a gas smell or a dog that cannot breathe is worse than no receptionist. Tell it which situations are emergencies, have it tell the caller to contact emergency services where that applies, and route anything urgent to a phone that someone actually carries.

Callers who will not talk to a bot. Some people hang up at the first synthetic word. Always offer a way out, such as a transfer, a message or a call back.

Privacy. A receptionist collects contact details and sometimes health or legal information. Ask each vendor where data is stored, for how long and how to delete it.

Where does PepoChat fit as an AI receptionist?

PepoChat is a website chat agent: it answers visitors from your own content, takes their details, books from a real calendar and hands conversations to your team. It is not a full AI phone receptionist today, because its optional voice feature runs on a separate assistant that does not share the chat agent's knowledge or inbox.

On the website, it covers all five receptionist jobs:

• Answers from your content. You import your website (a crawl of up to 200 pages counts as one knowledge source), upload PDFs and documents, and add Q&A pairs. The agent searches that knowledge base before every reply, and when it finds nothing it says so and offers a person instead of guessing.
• Takes messages. Lead capture asks for an email, plus name, phone or company if you choose, at the moment you pick: before the first message, when the agent needs to follow up, or just before a handoff. Everyone who shares an email lands on the Contacts page, and new leads can trigger a Slack card or a webhook on every plan, or an email to your team on Starter and Growth.
• Qualifies. Agent instructions let you tell it which questions to ask and what never to promise, in plain sentences.
• Books. It connects to a PepoSmart scheduling calendar with an API key. You choose which appointment types it may offer and write a "when to offer this" note for each. It shows open slots in the visitor's own time, books in yours, and the calendar re-checks the slot so nothing is double booked, then sends the confirmation and reminders.
• Routes to a person. It escalates when a visitor asks for a person, when it cannot help, or when something fails. Every member of your workspace gets an email with the last messages and a link, and the conversation waits in the team inbox with the full transcript. Business hours make it tell visitors when you are back, and a "no one answers live chat" mode suits a solo owner who replies by email.

The voice side needs an honest description. PepoChat can add a "Start a voice call" button and a "Call us" phone number to the widget, powered by your own Vapi account. You connect it with your Vapi keys, then pick an assistant and a phone number in the widget settings. The voice assistant itself is configured on Vapi's platform, and Vapi bills you for the minutes. It does not yet share the chat agent's knowledge base or tools, and voice calls are not saved in your team inbox. In practice that means two assistants to keep up to date and two places to look. The voice setup docs list exactly what visitors see.

So if most of your enquiries are phone calls, PepoChat on its own is not your AI receptionist yet; pair a phone product with it, or build the voice side on Vapi knowing it is separate. If most of your enquiries start on your website, it does the receptionist's job there, and every feature apart from new-lead emails to the team is on the free plan.

What to do next

Pull last month's call log and chat history, count where your enquiries arrive, and pick the channel that carries most of them. Then write down your dozen most common questions and their answers; you will need that list whichever product you choose.

If your website is where customers start, you can create a free PepoChat workspace, import your site, connect a calendar and test it in the dashboard before it goes live; the pricing page shows what each plan includes. If phone calls dominate, read the guide to AI voice agents for customer support before you sign up for any phone product, and run a two-week trial with real calls.