Skip to main content
CrewLogic Labs
Back to case studiesAI / Automation

Voice AI Receptionist — Answers, Books, and Knows When to Get a Person

A voice assistant (Vapi) backed by an n8n workflow that picks up a salon's calls, says up front that it's an AI, answers only from approved salon facts, books into the real Cal.com calendar, and hands complaints, reactions, and anything it isn't sure about to a person, then logs every call with its outcome and cost.

Vapin8nCal.comAirtableGemini AIResend
Tested with real voice calls: a real booking, grounded answers, and a handoff in under 2 seconds

The Problem

A salon's phone rings while everyone's hands are busy, and a missed call is usually a booking that goes to the place down the road. Voice AI can pick up, but most demos let the model improvise prices and book anything it's asked to, including the caller who just described an allergic reaction.

Our Approach

  • 1The assistant can only state facts from an approved block (hours, location, services, prices). Anything else becomes an offer to have the team call back, so it never invents an answer.
  • 2Bookings are checked in code before they reach the calendar: a real open slot, a listed service, and a time in the future. A keyword override turns any booking that mentions a reaction, injury, refund, or complaint into a handoff, whatever the model decided.
  • 3Real voice testing found the first booking took 14 seconds because logging ran before the reply, which made the assistant fill dead air twice. The workflow now answers the caller first and logs afterwards.

What This Achieves

  • A caller can ask for "Monday", hear three real open times, and end the call with a confirmed Cal.com booking.
  • A caller describing a reaction to hair dye is handed to the salon with their number and what they said, and is never booked in.
  • Every call becomes one Airtable row with its outcome, duration, and cost, so recovered bookings and cost per booking are real numbers.

Share this case study

Next Automation Case Study

Invoice Processing — AI Reads, Rules Check, a Person Approves

Have something similar in mind?

Scope Your Project