How to automate WhatsApp responses without a generic chatbot
By Marai ·
You have probably been there: you message a business on WhatsApp and get a numbered menu in reply. “Press 1 to book, 2 to cancel, 3 to speak to an agent.” You type “1”, it asks for the service. You type something it does not recognise. It drops you back at the menu. Three minutes in, you have lost patience and are reaching for the phone.
That kind of generic chatbot does not just fail to help — it actively harms. The client associates your business with a frustrating, impersonal, and ultimately pointless experience.
But automating WhatsApp responses does not have to be like that. In 2026, artificial intelligence makes it possible to build bots that understand what the client is actually saying — in their own words, in their natural way of writing — and resolve their request without feeling like a machine.
The problem with traditional chatbots
Rule-based chatbots work through decision trees. Each user response takes the conversation down a predefined branch. If the user says something that does not fit any branch, the bot breaks.
This creates specific problems:
Extreme rigidity. If the bot expects “1” and the user types “I want to book”, it understands nothing. The user has to guess what format the bot is expecting.
Impersonal experience. Every client gets the exact same sequence of questions, in the same order, with the same tone. There is no adaptation to context.
Frustration spirals. Every time the bot fails to understand and dumps the user back at the main menu, frustration builds. After two or three failed attempts, the client gives up.
No access to real data. Most generic chatbots are not connected to your calendar. They can collect information, but they cannot check real availability, confirm a booking, or process a cancellation.
How a natural language bot works
A bot built on natural language understanding (NLU) does not look for exact keywords. It analyses the full message and extracts what the user actually wants.
A real example
A client writes: “Hi, do you have anything free on Friday afternoon for a cut and colour?”
A generic chatbot would not understand this message at all. An NLU bot extracts:
- Intent: book an appointment
- Service: cut and colour
- Preferred date: Friday
- Time preference: afternoon
With that information, the bot checks your live calendar, looks for available slots on Friday afternoon for the “cut and colour” service, and presents the options directly:
“On Friday we have 4:00 pm with Laura or 5:30 pm with Marta. Which would you prefer?”
The client replies “4 o’clock with Laura” and the appointment is confirmed. No menus, no numbers, no repeating information.
What the bot needs to understand
For the experience to work in a service business, the bot needs to handle these intents:
- Book a new appointment — variations like “I’d like to make an appointment”, “do you have a free slot?”, “I need to see someone”
- Check existing appointments — “when is my next appointment?”, “what time was my Thursday one?”
- Cancel or reschedule — “I can’t make Tuesday”, “can I move to a different day?”
- Ask about prices — “how much is a facial?”, “price of a massage”
- Ask about opening hours — “are you open on Saturdays?”, “what time do you close?”
- General conversation — greetings, thank-yous, farewells
And it needs to manage context: if the client has already mentioned the service they want, the bot should not ask again.
What makes a bot connected to your calendar different
The key is not just language understanding. It is the connection to your business’s real data.
Real-time availability
When the bot offers slots, they are real slots. If 5 pm is already booked, it will not offer it. If a stylist has the day off, they will not appear. This eliminates the classic problem of bots that collect information and then require someone to manually validate it.
Instant confirmation
The client gets confirmation the moment they choose a slot. There is no intermediate step of “we will confirm shortly”. The appointment appears in your calendar automatically.
Integrated reminders
Once a booking is confirmed through the bot, automatic reminders are scheduled: 24 hours before, 2 hours before, according to your settings. The same WhatsApp channel handles both booking and reminders.
Client history
If the client has visited before, the bot knows. It can suggest the same service they booked last time, the same professional, and a similar time slot. The more they use it, the more personalised the experience becomes.
When the bot should hand off to a person
A good bot knows when it cannot resolve something and passes the conversation to a real person. This happens when:
- The client explicitly asks for it — “I want to speak to someone”
- The bot fails to understand after two or three attempts — better to hand off than to frustrate
- The query is complex — complaints, medical situations, payment issues
- The client is showing frustration — all-caps messages, negative language
The handoff should be transparent: “I am not sure I can help with this. I will pass you to the team.” And the entire conversation history should be visible to the person taking over, so the client does not have to repeat a word.
Setting up an AI bot with Marai
Marai’s WhatsApp bot is built specifically for service businesses. It is not a generic chatbot adapted for the purpose — it is a system purpose-built for appointment management.
What it does
- Understands messages in natural language (Spanish, including slang and abbreviations)
- Checks real-time availability in your calendar
- Allows booking, modifying, and cancelling appointments
- Handles multiple services and professionals
- Sends instant confirmation
- Schedules automatic reminders
- Hands off to a human when necessary
What it does not do
- It does not invent information about your business
- It does not confirm appointments in slots that do not exist
- It does not send unsolicited promotional messages
- It does not store data outside your system
How to configure it
Configuration is straightforward: activate the bot from the Marai dashboard, define which services are available for WhatsApp booking, which professionals are handling them, and during which hours. The bot uses that information to manage conversations. You do not need to build flows or write predefined replies.
The balance between automation and the human touch
The goal is not for the bot to replace all human interaction. It is for the bot to handle the repetitive parts — the tasks that require no judgement or empathy — so that you and your team can focus on what actually needs those qualities: serving the clients in front of you, resolving complex situations, and creating experiences that earn loyalty.
A well-implemented bot does not dehumanise your business. It sets it free.
Explore Marai’s automation features →
Learn about the WhatsApp bot in detail →
Start for free → — AI WhatsApp bot available from the Starter plan (€29/month).