How an AI receptionist books appointments by phone

Most people who hear "AI receptionist" for the first time picture either a chatbot (text on a screen) or the menu-driven phone trees they remember from 2005 ("press 1 for hours…"). The actual thing is neither. It's a real-time conversation, one that opens by telling you it's an AI and then sounds a good deal less robotic than that sentence suggests.

Here's what actually happens during one call, start to finish and with no jargon.

The call, step by step

It's 8:47pm on a Tuesday. Sarah has a dripping kitchen tap and calls Mike's Plumbing. Mike is at his daughter's hockey game.

1. Pickup

Mike's phone is on silent at the rink, so the call goes to Cedri, which answers right away in the voice Mike picked, with his greeting: "Hi, you've reached Mike's Plumbing. I'm an AI assistant and this call may be recorded. How can I help? Si vous préférez le français, dites-le-moi."

2. Understanding the request

Sarah says "my kitchen tap won't stop dripping, could someone come this week?" Cedri picks out what she needs (a tap repair), when (this week) and that she wants an appointment.

Had Sarah said water was pouring into her basement, Cedri would not book a slot at all. For plumbers and other trades with real emergencies, it takes the address and a callback number and puts the call straight through to the number Mike set, or takes an urgent message if he can't pick up.

3. Checking the calendar

Still on the call, Cedri checks Mike's agenda in Cedri, and his Google Calendar too, since he connected it (calendar sync is in beta). It finds the first open slot that fits: "I can have someone there Wednesday at 10 am. Does that work for you?"

4. Confirming details

Sarah agrees. Cedri asks for her address, her phone number for the confirmation SMS, and a one-line description of the problem to help Mike prepare. None of that feels like a form. It is just the rhythm of a normal phone conversation.

5. Booking and confirmation

Cedri creates the booking and texts Sarah: "Confirmed: Your appointment with Mike's Plumbing is booked for Wednesday at 10:00 AM. Reply 2 to cancel. Mike's Plumbing". When the call ends, Mike gets a text with Sarah's number, "Appointment booked!" and a one-line summary of the call.

Sarah hangs up with a real appointment. Mike sees the booking notification at intermission, knows what to bring, and goes back to the game.

Total time on the call: about a minute. Time Mike spent on it: none. He didn't know it happened until the text arrived.

The pieces under the hood

There's no magic here, just three well-understood technologies wired together:

• Speech-to-text + text-to-speech. The modern models Cedri uses process audio as the caller speaks rather than waiting for them to finish. That is what removes the long pauses and flat delivery which used to give a phone bot away.
• A language model. It gets your business's details (services, hours, prices, FAQs) and is told to stick to them and never make things up. It can only do a short list of things: check times, book, move or cancel a booking, answer from your FAQs and website, and pass the call to you.
• Calendars. Every booking goes into Cedri's own agenda. If you connect Google Calendar or Outlook (in beta), Cedri also checks it for busy times during the call and copies the booking into it. The caller gets a confirmation text. You get a text after the call, plus an email and an app alert if you keep them on.

What it doesn't do

Worth being honest about the boundaries. An AI receptionist is not:

• A general-purpose assistant. It answers calls to your line and texts sent to it. It doesn't sort your email or reply on social media.
• A replacement for human judgement on hard calls. When a call needs a person (a real emergency, a caller who asks for you, a question it can't answer), it puts the call through to your number, or takes a message if you can't pick up.
• A magic upsell engine. It books what people ask for. It doesn't aggressively sell add-ons. (If you want that, you want a sales team, not a receptionist.)

Why this works now and didn't two years ago

Two things changed. Voice models got fast enough that the long pauses are gone, and language models got reliable enough at narrow tasks ("only book real open slots") to trust with real customers.