Automating customer bookings: 7 steps from first message to confirmed appointment

Before connecting anything, open a spreadsheet and list three columns: service name, duration in minutes, price.

For a hair salon: "Women's haircut — 60 min — $45." For an auto shop: "Oil change — 45 min — $65." If you have 5 providers, add a "who performs" column — the assistant needs to know that "lash extensions" is only done by Sarah, not any available technician.

Common mistake here: padding durations "just in case." The customer only sees the end time in the schedule, and you lose slots. Write the real duration and configure a buffer separately in the system.

Each provider (or service, or room — depending on your business logic) gets a separate Google Calendar. This is the standard that works with any CRM and most AI assistants, including ZiFlow.

ZiFlow connects calendars via OAuth — click "connect," pick the Google account, grant access. In 30 seconds the assistant sees availability in real time.

One important note: the calendar has to be the work calendar, not a personal one. If your provider puts birthdays and doctor visits on the same Google Calendar, the assistant treats those as booked slots. The fix: a dedicated work calendar like "Sarah — Work" per provider.

The assistant needs to know not just when you're open, but also lunch breaks, each provider's days off, and which holidays close the business.

This is especially critical around Thanksgiving, Christmas, July 4th, or whatever dates matter in your market. Forget a holiday, the assistant will offer slots on a closed day, the customer will confirm, show up — and leave angry.

In ZiFlow, hours are set at the business level (general schedule) and at the provider level (individual). The provider takes priority: if the salon is open but Marco has a day off, his slot simply isn't offered.

This is the core of the automation. When a customer messages in, the assistant needs to collect: service, preferred time, provider (if it matters), contact phone, name.

A good conversation structure:

1. Greeting + clarify what they want.
2. Offer 2-3 slots from the nearest available.
3. Confirm name and phone (if new customer).
4. Final confirmation with date, time, service, price.
5. Create the event in Google Calendar + log the lead in CRM.

The most important thing: don't ask 10 questions in a row. The assistant should run the conversation like a person — short turns — and pull info from context when possible. "Want to come in tomorrow at 6 for a cut with Sarah" is already 4 out of 5 parameters.

This is where "just booking" splits from "booking with deposit." For services with high cancellation rates (medspa, expensive procedures), a 20-30% deposit reduces no-shows by 60-70%.

ZiFlow integrates with Stripe and local processors like FreedomPay — the assistant sends a payment link right in the chat, the customer pays in one tap, and the "paid" status auto-fills the lead card. Without a deposit, the appointment is created in "pending confirmation" status.

Don't require a deposit for every service — that will kill 30-50% of inbound. Apply it selectively: expensive procedures, VIP providers, or customers who've already no-showed once.

The simplest and most underrated step. One reminder 24 hours before the visit cuts no-shows from an average 22-28% to 12-15%.

ZiFlow sends reminders automatically in the customer's time zone. The text is customizable — I recommend keeping it short with easy cancel:

"Hi! Reminder: tomorrow at 6pm you're booked with Sarah for a haircut. If your plans change, just reply 'cancel' or 'reschedule' and we'll sort it."

The key phrase is "if plans change, reply." The customer sees that cancellation is easy, so they actually cancel in advance instead of just not showing up. You get time to fill the slot.

The first 7 days aren't "works / doesn't work" — they're data collection. What to watch:

• How many bookings completed fully automatically (no staff involvement).
• At which step the assistant hands off to a human — that's a signal where a scenario is missing.
• How many no-shows — compare to pre-automation baseline.
• Open-text complaints from customers — "bot didn't get it," "awkward" — go straight into the improvement backlog.

In a well-tuned system after 3-4 weeks, 80-90% of bookings are fully automatic. Staff only steps in for the 10-15% of complex cases (reschedules, group bookings, VIP).