The AI receptionist onboarding checklist: what to set up before your first live call

Most AI receptionist setups that disappoint do not fail because the technology is bad. They fail because the business turned the system on before it was ready, and the first real callers got wrong answers, missed bookings, or awkward silences. An onboarding checklist fixes this. It forces you to answer every question the AI will need before a caller asks it.

This checklist covers the twelve setup steps that matter, in the order that makes sense. Work through it once, and your first live call will sound like the hundredth.

One: define what the AI is allowed to do

Before touching any settings, write down the AI’s job description in plain language. Answer calls, book appointments, take messages, answer pricing questions, transfer emergencies. Also write down what it may not do: quote custom prices, cancel existing appointments, give medical or legal advice. This list becomes the foundation for everything else, because every setting below exists to enforce it.

Two: choose and configure the phone number

Decide whether the AI answers your main business number, a new dedicated number, or specific lines. If it answers the main number, set up call forwarding from your carrier or phone system. Test the forwarding with a real call from your cell phone before going further, because forwarding misconfigurations are the most common reason an AI “isn’t working” on day one.

Three: write the greeting

The greeting is the first thing callers hear, and it sets the tone for the whole call. Keep it short: your business name, a welcome, and what the caller can do. “Thanks for calling Maple Dental, how can I help you today?” beats a thirty-second monologue about your mission statement. Record or select the voice, then listen to it on a phone call, not just in the dashboard preview. Voices sound different over a real phone line.

Four: build the knowledge base

The AI can only answer questions you have taught it. Gather your hours, location, services, pricing, policies, and the ten questions your staff answers most often. Enter them in the knowledge base in the caller’s language, not yours. Callers ask “how much to clean a three-bedroom house,” not “what is your rate card for residential properties.” Short, specific entries work better than long documents.

Five: connect the calendar

If the AI will book appointments, connect your calendar now and configure buffers, lead times, and working hours. Book a test appointment through a test call and confirm it lands on the calendar correctly. Calendar setup deserves its own session, because it is where most booking errors originate.

Six: connect the CRM or booking system

If call data should flow into HubSpot, GoHighLevel, or your scheduling tool, connect it now and map the fields. Decide which call outcomes create contacts, which pipeline stages they land in, and who owns the follow-up. Test with your own number so you can see exactly what the CRM record looks like.

Seven: set the business hours and after-hours rules

Tell the system when you are open, when you are closed, and what should happen in each case. After-hours calls might get booking-only handling, or message-taking, or emergency triage. Holiday schedules go in too. If you skip this step, the AI treats 2:00 a.m. like 2:00 p.m., and callers get offered appointments at times nobody works.

Eight: configure call routing and escalation

Decide when the AI transfers a live caller to a human. Common triggers: the caller asks for a specific person, the caller is upset, the question is outside the knowledge base, or the caller explicitly asks for a human. Set the transfer numbers and test each one. A transfer to a number that rings forever is worse than no transfer at all.

Nine: set up notifications

Decide what you hear about and how. Most businesses want an instant text for booked appointments and urgent calls, and a daily or per-call email summary for everything else. Too many notifications and you will start ignoring them, which defeats the purpose. Too few and you will miss the call that mattered. Start with more, then trim after the first week.

Ten: configure spam filtering

Turn on the platform’s spam and robocall screening, and set the rule for unknown numbers. Most businesses let unknown numbers through but flag them, because a new customer is by definition an unknown number. Review the blocked-call log in the first week to make sure real callers are not getting caught.

Eleven: run the test call matrix

Place at least five test calls covering the scenarios your business actually gets: a straightforward booking, a pricing question, an after-hours call, a caller who asks for a human, and a caller who goes off script. Listen to every recording. Fix what sounds wrong. This is the step that separates good setups from bad ones, and it is the step most often skipped.

Twelve: plan the first week of monitoring

For the first week, review every call summary daily. You are looking for wrong answers, missed bookings, and questions the knowledge base could not handle. Add the missing answers to the knowledge base as you find them. Most setups need two or three rounds of these corrections before they settle. After the first week, drop to a weekly review, then monthly.

What to have ready before you start

The checklist goes faster if you gather materials first: your business hours including holidays, your service list with prices, your cancellation and refund policies, your staff directory with transfer numbers, your calendar login, your CRM login, and the answers to your ten most common caller questions. With those in hand, most businesses complete the full checklist in an afternoon.

The single best predictor of a smooth launch is step eleven. Businesses that test thoroughly before going live have quiet first weeks. Businesses that skip it spend the first week apologizing to callers. If you only have time for one step, make it the test calls. For a deeper testing plan, see how to test your AI receptionist before launch, and keep the knowledge base guide nearby while you work through step four.