WhatsApp + AI booking

WhatsApp AI appointment booking: real slots, customer details and a clean handoff to sales

Yes, an AI agent can answer WhatsApp questions about available time slots, collect the customer's details, book the appointment and pass serious inquiries to your sales team with the conversation attached, and it can do it reliably without sounding like a robot. That is WhatsApp AI appointment booking: an agent on your company number that reads your real calendar, books, reschedules and cancels, sends reminders, and knows when to hand off to a person. It is for clinics, studios, academies and service businesses that take bookings on WhatsApp and answer by hand, often after hours. I build it as an independent AI systems engineer (Sr DevOps, ten years of production infrastructure) on Meta's official Cloud API, connected to Google Calendar, Cal.com or your booking system. I run one in production today for a small business that sells sessions on fixed schedules. Three-week sprint, a single fixed price agreed up front, and a 60-day operational guarantee.

How does a WhatsApp AI agent know which time slots are actually free?

It asks your calendar, every time. When a customer writes "do you have anything Thursday afternoon", the agent calls the availability endpoint of the system you already use, Google Calendar, Cal.com or the API of your booking software, and answers with the slots that are free at that moment. It never works from a list of hours written into its instructions, because that list is wrong the first time someone books by phone, a staff member takes a day off or you block a morning for a meeting.

Availability rules stay where they already live. If your booking tool knows that a first consultation takes 60 minutes, that Dr. Pérez does not work Fridays and that you need a 15-minute buffer between sessions, the agent respects all of it because it reads the result instead of recalculating it. For businesses with fixed schedules (classes, group sessions, workshops) the same applies: the agent reads which sessions have places left and offers those. When the calendar cannot be reached, it says so plainly and offers to have a person confirm, instead of guessing.

Live lookup on every question, never a hard-coded list of hours
Durations, buffers, staff and resources come from your booking tool
Offers two or three concrete options instead of an open question
Time zones handled explicitly for customers writing from abroad
If the calendar is unreachable, it escalates instead of inventing a slot

Can it book, reschedule and cancel, and collect customer details first?

Yes. Booking is a write to the same system: the agent creates the event or the booking through the API, and the confirmation the customer receives contains the date, time, place or video link and how to change it. Rescheduling and cancelling work the same way, with your rules applied. If cancellations within 24 hours carry a fee or need approval, the agent explains the policy and routes the request to a person instead of deciding on its own.

Before it holds a slot, the agent collects what your team needs to run the appointment: name, email, the service, and the two or three questions you always ask (for example, first visit or returning client, the reason for the visit, the company name for a B2B call). It asks them conversationally, one or two at a time, and skips what it already knows from earlier messages or from your CRM. The answers land in the booking itself and in your CRM or sheet, so nobody re-asks the customer at the door.

Creates, moves and cancels bookings through your calendar or booking API
Collects name, contact and your intake questions before confirming
Applies your cancellation and rescheduling policy, escalates exceptions
Writes the details to the booking and to your CRM or sheet
Sends a payment link from your existing provider when a deposit is required

How do appointment reminders work on WhatsApp, and what is the 24-hour rule?

When a customer writes to your business, a customer service window opens (24 hours at the time of writing) in which you can reply freely. A reminder for an appointment three days away will almost always be sent outside that window, so it has to go out as a message template that Meta has reviewed and approved, and only to customers who agreed to receive messages from you. Templates have fixed text with variables (name, date, time, address), and Meta prices them by category.

The usual pattern is two reminders: one 24 hours before, with buttons or a short instruction to confirm, reschedule or cancel, and one 2 hours before with the address or the video link. When the customer answers the reminder, a new service window opens and the agent handles the reply in free text, including "can we move it to Friday", which it resolves against the live calendar. I run the same pattern for my own meetings: an n8n workflow reads Google Calendar and sends reminders by email and WhatsApp at 24 hours, 2 hours and 15 minutes before each call. Meta's policies and prices change, so before each build I check the current documentation.

Can the agent handle after-hours inquiries and pass serious ones to sales?

After hours is where the agent earns most of its keep. A customer who writes at 22:00 asking about prices and availability gets an answer from your price list and a real slot in the calendar, instead of waiting until the next morning and booking with someone else. Questions it can answer from your documents (services, prices, preparation, location, policies) are answered at any hour. Questions it cannot answer are acknowledged honestly, with a clear promise of when a person will reply.

Serious inquiries go to sales with context, not as a forwarded chat. You define what "serious" means: a company asking for a group or corporate booking, a budget above a threshold, someone ready to pay, a returning client, a complaint. When one of those rules fires, the agent pauses on that thread, notifies the right person in Slack, email or WhatsApp, and attaches a short summary: who the person is, what they want, what they already said, and what the agent promised. During business hours that is a live handoff. After hours it becomes the first item on the queue in the morning, already qualified.

Answers prices, services and availability at any hour from your sources
Rules you define decide what counts as a serious inquiry
Summary and full history attached to every handoff
The agent pauses on the thread while a person handles it

Can a WhatsApp booking agent sound non-robotic?

Yes, and it mostly comes down to what it is built from. Before writing any instructions I read a sample of your real conversations, anonymized, to learn how your business actually talks: the greeting, the level of formality, the words your customers use, the way your team explains prices. The agent writes in that voice. It sends short messages, one idea at a time, the way people write on WhatsApp, instead of a paragraph with bullet points and three emojis.

The other half is knowing when to stop. A robotic bot keeps pushing the next step; a good agent answers the question that was asked, offers the booking once, and leaves it there if the customer is just browsing. It does not repeat the same sentence when it is stuck, it hands off. It does not pretend to be a person either: if someone asks, it says it is the business's assistant and that a person is one message away. In practice customers care far more about getting a correct answer quickly than about who typed it.

What does a reliable WhatsApp booking agent look like under the hood?

Reliability starts with the channel. The agent runs on the WhatsApp Business Platform through Meta's official Cloud API, not on an unofficial bridge that automates the phone app and gets numbers banned. In the production agent I run today, the business keeps its existing number and keeps using the WhatsApp Business app on the phone in coexistence mode, so the team can still read and reply by hand while the agent works through the API.

In plain words, the build is this. Trigger: a message arrives at a webhook Meta calls, which verifies the signature, deduplicates and queues it. Runtime: an n8n workflow or a small Python service loads the conversation memory for that number. Model: Claude or OpenAI through their APIs, with versioned instructions about tone, what it may promise and when to escalate. Tools: availability, booking, rescheduling, CRM writes, payment links. Guardrails: no invented slots or prices, approval for exceptions, a kill switch per thread and global. Every message, tool call and handoff is logged in your own database, failed calls are retried, and an alert fires when something keeps failing. It runs in your Meta account and your hosting, and you own the code, prompts and data.

Official Meta Cloud API, coexistence with the phone app where available
Webhook with signature verification, retries and deduplication
Calendar and booking APIs as the single source of availability
Full audit log and alerts in your own infrastructure
You own every account, the code and the prompts

How much does a WhatsApp AI appointment booking agent cost?

The reference build is the AI Intake Agent with knowledge base, a productized scope quoted at a fixed price, with an optional monthly operation for iteration after the guarantee period. That covers the scope on this page: the channel set up, answers from your documents, live availability, booking with intake questions, reminders through approved templates, handoff to sales, and the audit log. Larger scopes, for example several locations, several numbers or quoting on top of booking, are quoted as a single fixed price after a discovery call, with the AI Operating System as the upper reference point.

What is not inside the fixed price: the messaging fees Meta or a provider charges you, the model API usage and the hosting. All of those are billed to you directly, in your name, with a cap declared in the proposal, so there is no markup on tools. I do not bill hourly for builds and I do not resell software.

When is your booking tool's own WhatsApp reminder enough?

Many booking tools already send confirmations and reminders, and some send them on WhatsApp. If your customers book through a link, your problem is no-shows, and nobody needs to talk before or after the booking, turn on those reminders and you are done. If you get a handful of messages a day and they are all "what are your hours", the greeting and quick replies in the WhatsApp Business app solve it for free. In both cases I will tell you so on the first call, because I do not resell tools and a custom agent would be money badly spent.

An agent earns its cost when booking happens inside a conversation: customers ask questions before they commit, want a specific time, need to reschedule by chat, or write after hours when nobody is there to answer. It also earns it when a share of those conversations are worth real money and should reach a salesperson fast, with context. If that sounds like your inbox, the 15-minute call is where we check it against your real numbers.

Frequently asked questions

Can AI answer WhatsApp questions about time slots and pass serious inquiries to sales?

Yes. The agent reads your live calendar to answer availability questions, collects the customer's details, books the slot and, when a rule you define fires (a group booking, a high budget, a complaint), hands the thread to a salesperson with a summary and the full history. It runs on Meta's official Cloud API and logs every step, which is what makes it reliable.

Which calendars and booking systems does it work with?

Google Calendar and Cal.com are the most common, and any booking or scheduling software with an API can be connected. The agent reads availability from that system and writes bookings back to it, so your team keeps working in the tool they already use. If your system has no API, we discuss alternatives on the first call before anything is quoted.

Can I keep my WhatsApp number and keep using the app on my phone?

Often yes. In coexistence mode the number stays registered on the WhatsApp Business app and is also connected to the Cloud API, so you keep reading and answering on the phone while the agent works. I run an agent in production this way. Availability depends on Meta's current rules for your country and account, which I check in the first week.

Can the agent send reminders 24 hours and 2 hours before the appointment?

Yes, as approved message templates to customers who opted in, since reminders usually fall outside the 24-hour customer service window. When the customer replies to a reminder, the agent handles the answer in free text, including rescheduling against the live calendar or cancelling under your policy.

Will customers know they are talking to an AI?

If they ask, yes. The agent writes in your business's voice, learned from your real conversations, with short messages and no menus, but it does not pretend to be a person. It says it is the business's assistant and that a person can take over at any moment, and it hands off when asked.

What can the agent not do?

It does not take card payments inside the chat; it sends a link from your payment provider. It does not give clinical, legal or financial advice, and it does not decide exceptions to your policies; those go to a person. It cannot message people who never opted in, and I do not build cold outreach on WhatsApp.

Fifteen minutes on the bookings your phone handles by hand.

A free call. I'll tell you straight which processes today's AI can solve and what your infrastructure needs for them to actually run. If it fits, we move forward. If not, I point you the right way, free.