AI Receptionist Agent for Local Businesses: Calls, Bookings and Reviews

1. The missed-call problem this agent solves

For a local business, the front desk is a single point of failure. The stylist is mid-haircut, the dentist is mid-procedure, the plumber is under a sink — and the phone rings into voicemail, where most callers simply hang up and dial the next business on the list. An AI receptionist agent removes that bottleneck: it answers chat and calls instantly, books real appointments into the real calendar, takes structured messages outside business hours, and asks satisfied customers for the reviews that keep the business visible.

This guide builds that receptionist. It is Guide 2 of our AI agents series; if you have not seen the model overview, start there — and note that our existing B2B voice receptionists blueprint covers the sales packaging side of this same agent in depth.

2. What the receptionist does, hour by hour

• Answers instantly, on chat and on calls, with the business's services, prices-as-published and policies — and nothing it invents.
• Books, reschedules and cancels appointments against the live calendar: real free slots only, with confirmations and reminders sent automatically.
• Applies business-hours logic: inside hours it can offer a warm transfer to staff; outside hours it books what it can, takes structured messages for the rest, and never promises a human "in five minutes" at midnight.
• Requests reviews at the right moment — after a completed appointment, once per visit, with a direct link — and never nags a customer who had a problem; those route to the owner instead.
• Reports daily: a summary of bookings, messages and unresolved questions lands in the owner's inbox, so the human stays in charge of a system they did not have to babysit.

3. The stack: voice/chat agent + calendar + rules

Four roles again. For the voice channel, use a voice-agent platform that connects a speech recognizer, an LLM and a natural voice to a phone number (Retell AI and Vapi are the platforms our voice-agents blueprint builds on); for chat, the same brain can sit behind the website widget and the business's messaging inbox. Behind both channels:

• The calendar is the source of truth (Google Calendar or the booking system the business already uses). The agent reads free/busy and writes confirmed bookings — it never keeps its own parallel schedule.
• A business profile document: services with durations, staff, hours, holidays, cancellation policy, and the exact facts the agent may state. Every wrong answer you ever fix goes into this document.
• Business-hours and handoff rules as configuration, not vibes — section 5 lists them.
• A review link and a completion trigger: when an appointment is marked completed, the workflow waits a defined interval and sends the review request, once.

4. Configuration steps, in order

1. Write the business profile with the owner: services, durations, prices exactly as published, hours, and the ten questions customers ask most — answered in the owner's words.
2. Connect the calendar read-only first. Test that the agent quotes only genuinely free slots; then enable writes on a test calendar before touching the real one.
3. Set the greeting and identity: the agent states it is the business's virtual assistant. Hiding that it is an AI buys nothing and costs trust the first time a caller notices.
4. Wire the hours logic: open / closing-soon / closed behaviors, holiday table, and what gets booked automatically versus taken as a message.
5. Configure the review request: trigger on completed appointment, single send, exclusion list respected forever.
6. Run the test script: twenty scripted calls and chats — accents, background noise, a kid shouting, the questions owners dread — and read or listen to every transcript before launch.

5. Example dialogues

A simple booking inside business hours:

Customer: Hi, do you have anything open Saturday morning
          for a haircut?
Agent:    We do. Saturday has 9:30 and 11:00 open for a cut
          (45 minutes). Which works better for you?
Customer: 9:30 please. It's for Laura.
Agent:    Booked: Saturday 9:30, haircut, for Laura. I have sent
          a confirmation and a reminder the day before. Anything
          else?

And the after-hours case, where the agent's job is to resolve, not to promise a human:

Caller (11:40 p.m., salon closed): Hey, I need to move my
          appointment tomorrow, something came up at work.
Agent:    Of course. Your appointment tomorrow at 10:00 with
          Marta — I can move it to Thursday 10:00 or Friday
          17:30 with the same stylist. Which do you prefer?
Caller:   Friday 17:30.
Agent:    Done: moved to Friday 17:30, same service and stylist.
          Confirmation sent. The salon opens at 9:00 if you
          need anything else in the morning.

6. Handoff-to-human rules (the section owners actually buy)

The receptionist is trusted because of where it stops. Fix these rules in writing with the owner before launch:

• Instant handoff: the caller asks for a person, sounds distressed, mentions an emergency, a complaint, a refund, or anything legal or medical. The agent stops, takes a structured message, and alerts staff immediately during hours.
• Two-strike rule: if the agent fails to understand twice, it does not try a third time — it apologizes, takes a message and flags it for callback.
• Never-guess rule: prices, availability and policies that are not in the business profile are confirmed by a human, not improvised.
• After-hours promise rule: the agent may book and message, but may only promise response times the owner approved in writing ("we reply by 9 a.m.").

7. Example monthly-fee model (example math, not an earnings claim)

As in Guide 1, this is an illustrative example with labeled assumptions — check the shape, then recompute with your own figures:

Under those assumptions, one business contributes roughly $160–$200 per month after running costs; four businesses roughly $640–$800. The risk variable is usage: a single very busy phone line can eat the margin in voice minutes, which is why the fee model needs a fair-use note and why you monitor minutes weekly in the first month. Model your own numbers with the AI agent cost calculator.

8. Mistakes that get receptionists fired

• A second, private calendar. Double bookings are unforgivable; the business calendar is the only truth.
• Pretending to be human staff. One discovered deception outweighs a hundred smooth calls.
• Review-begging everyone. Asking an unhappy customer for a public review — instead of routing them to the owner — turns a service tool into a reputation hazard.
• No transcript review routine. The agent's mistakes are all in the logs, waiting to be read. Weekly, at least at the start.
• Unbounded small talk. A receptionist that chats for ten minutes is a cost center; politeness with a steering wheel wins.

Previous in the series: Guide 1 — the sales agent. Go deeper: the B2B voice receptionists blueprint for packaging and pitching, and the series hub for the full roadmap. Next in the series: Guide 3 — the content agent.