- An AI agent is not a button bot; it understands the customer's own words and replies in Saudi dialect using only your company's information.
- It can book, reschedule and cancel appointments in your real calendar and send confirmations and reminders.
- You decide what it does alone and what needs staff approval before it happens (exceptions or sensitive requests).
- If the customer asks for a person, or the question is outside its knowledge, it hands the chat to a human with a full summary.
Imagine a dental clinic in Riyadh receiving 200 WhatsApp messages a day: "How much is a cleaning?", "Any slot tomorrow afternoon?", "I need to move my appointment", "Where are you located?". Most questions repeat, many arrive after hours or at the weekend, and a customer who gets no reply within minutes messages another clinic.
That is where an AI agent on WhatsApp comes in: a digital team member that answers 24/7, understands the customer, books the appointment and knows when to hand over to a human colleague. This article explains how it works, how it differs from a classic bot, and how to set it up safely without losing control.
The idea is not to replace your team; it is for the AI agent to take repetitive questions and simple bookings off their plate and hand them the conversations that need expertise and a human decision.
Classic bot vs AI agent
A classic WhatsApp chatbot works with menus: "press 1 for appointments, 2 for prices". Great for clear paths, but it struggles when a customer writes freely, such as "I need a slot for my son after school on Sunday if there is one". An AI agent understands that sentence and acts on it.
| Item | Button bot | AI agent |
|---|---|---|
| Understanding | Fixed options and keywords | Free text in Arabic and English |
| Source of answers | Pre-written texts | Your knowledge base (services, prices, policies) |
| Booking | A fixed step-by-step path | Suggests free slots and books on request |
| Unexpected questions | Back to the menu or stuck | Apologises and hands over with a summary |
| Best for | Simple paths and menus | Varied conversations and bookings |
In practice many companies use both: the bot for welcome and quick routing, the AI agent for open conversations and booking.
How does it actually book?
A booking is not a promise typed into the chat; it is a real booking in Klamna appointments, the same calendar the front desk sees:
- The customer writes: "I'd like a cleaning next week, afternoon".
- The AI agent identifies the service and its duration and checks free slots for the right doctor or department.
- It suggests two or three options: "We have Monday 4:30 or Tuesday 5:00, which suits you?".
- The customer chooses; the agent confirms name and mobile and books.
- The customer gets a confirmation and a reminder before the visit, and the appointment appears on the customer's CRM record.
Rescheduling and cancelling work the same way: the customer writes "I can't make it tomorrow", and the agent finds the booking and offers alternatives. That reduces no-shows, because moving an appointment is easier than ignoring it.
Approvals: you set the limits
The biggest worry for managers is "what if it says something wrong, or gives a discount we never agreed?". That is why Klamna AI agents include approvals:
- Automatic actions: answering from the knowledge base, showing free slots and standard bookings.
- Actions needing approval: a booking outside normal policy, a sensitive request, or any action you choose. The agent prepares the request, a staff member approves or rejects with one click, and the agent continues with the customer.
- Clear boundaries: your instructions list what it must never discuss (for example medical diagnosis or prices that do not exist).
The knowledge base: it knows what you teach it
The AI agent does not invent facts about your company; it answers from the knowledge base you prepare: services, durations, branches and locations, working hours, accepted insurers, cancellation policy and FAQs. The clearer the knowledge, the more accurate the answers.
A practical way to build it:
- Collect the 30 most common WhatsApp questions from last month and write an official answer for each.
- Add your approved services list and prices (if you publish them) and the cancellation and rescheduling policy.
- Update it whenever something changes: a new offer, branch or doctor.
Handing over to a human: when and how
A good AI agent knows its limits. It hands the conversation to a person when:
- The customer explicitly asks: "I want to talk to someone".
- The question is not in the knowledge base.
- The customer is upset or has a complaint.
- An action needs a human decision.
Hand-over happens inside the same unified inbox; the agent receives the chat with a summary of what the customer asked, what the AI did and where it stopped, so nobody asks the customer to start again.
Examples by industry
| Industry | What the AI agent does |
|---|---|
| Clinics and medical centres | Books and reschedules, answers about services, insurance and locations, sends reminders |
| Salons and beauty centres | Books by service and specialist, shows free slots |
| Car service and maintenance | Books inspection or service slots, answers about service duration |
| Real estate | Qualifies the customer (city, budget, type) and books a viewing with a rep |
| Schools and training centres | Answers about programmes and enrolment and books a visit or interview |
For clinics specifically, see clinics and medical centres, with a ready template that builds appointments, the bot and customer fields from sign-up.
After hours, weekends and peak seasons
The AI agent proves its value most when nobody else is answering. A large share of customer messages arrives at night, on Fridays or on the first days of holidays. Customers used to wait until morning; now they get an answer in seconds and leave with a booked appointment.
In peak seasons such as Ramadan, Eid, back-to-school and seasonal offers, messages can double overnight. Instead of hiring temporary staff, the AI agent absorbs the extra repetitive questions while your team focuses on conversations that need a decision. You can also adjust its behaviour by time: full answering and booking after hours, quicker hand-over to staff during working hours.
Privacy and customer data
Conversations contain names, numbers and sometimes health information, so security is part of the design, not an add-on:
- The AI agent only sees what the conversation needs and never reveals one customer's data to another.
- Every conversation is saved on the customer record, so you can review exactly what the AI said.
- Permissions control which staff can edit the instructions and the knowledge base.
- For clinics: state clearly in the instructions that it never gives a diagnosis or treatment advice and passes medical questions to the team.
Measuring results
After the first month, compare these figures with the month before:
| Indicator | What to expect |
|---|---|
| First response time | Down from minutes or hours to seconds |
| Bookings made after hours | A new source of appointments instead of lost enquiries |
| Share of chats handed to staff | Shows what is missing from the knowledge base |
| No-shows | Fewer, thanks to reminders and easy rescheduling |
| Customer satisfaction | Measured with a short survey after the chat |
Review the handed-over conversations weekly: any question that appears three times goes into the knowledge base, so the agent improves week after week.
Going live in one week
- Day 1: prepare the knowledge base (questions, services, policies).
- Day 2: set up the calendar: doctors or staff, services and durations, working hours.
- Day 3: write the agent's instructions: tone, boundaries, when to hand over.
- Day 4: decide which actions are automatic and which need approval.
- Days 5 and 6: test internally with real questions and adjust.
- Day 7: switch it on after hours first, then extend to all hours.
Common mistakes to avoid
- An empty knowledge base: launching without enough information makes the agent hand everything to staff, so it feels useless.
- Vague instructions: "be polite" is not enough. Give examples of replies you like and replies you do not, and explain how to write prices and times.
- Hiding that it is AI: it is better for it to introduce itself as a digital assistant, with the customer free to ask for a person at any time. Clarity builds trust.
- An outdated calendar: if doctors' or staff leave is not in the calendar, it will book slots that are not available. Keep it up to date.
- Nobody watching approvals: assign someone to handle approval requests during working hours, so customers do not wait.
Want to see it answer and book on your number? Book a demo, or sign up, build your own package and see the price before you pay.
FAQ
Does it speak Saudi dialect?
Yes. It can reply in clear, polite Saudi dialect, in Modern Standard Arabic or in English, following the customer's language.
Can it give wrong information?
It answers from your knowledge base and instructions. Even so, we recommend more approvals and reviewing conversations at the start. Anything outside its knowledge goes to a person.
Does it work on the official WhatsApp API?
Yes. It works on Klamna inbox channels, first of all the official WhatsApp API, and on other connected channels too.
Does it answer voice calls as well?
The chat AI agent handles written conversations. For calls there is the AI voice agent, which speaks Saudi dialect.
Does it replace my front-desk staff?
No. It takes repetitive questions and simple bookings off their plate and covers after hours, so they can focus on cases that need a person.
Ready to let WhatsApp take bookings while you sleep? Talk to us or WhatsApp us on 966920031338.