An agent built for your business.On WhatsApp, Instagram, web chat and the phone.
It answers in Arabic and English, and it finishes the job.
Your customer messages at eleven at night. Nobody is on the phone, and the first reply lands at nine the next morning, if it lands. We do not sell the same agent to everyone. We build the one your business actually needs, on your process, your rules and your knowledge, and it works every channel you use from a single brain, so the answer on WhatsApp is the answer on the phone. It does not hand the customer a menu. It reads the voice note, checks what is actually available, books it, and confirms it. When it should not be the one answering, it stops and puts a human in.
Not a chatbot. An employee with a job to finish.
A chatbot answers a question and stops. An agent checks the calendar, writes the booking, sends the confirmation, and knows when to stand down. The difference is not how it talks. It is what it is allowed to do.
What it does
- Reads voice notes, photos and documents
- Checks live availability before it promises anything
- Writes the booking into your system
- Hands the conversation to a human and stays out of it
What it is not
- Not a keyword menu with an AI voice on top
- Not a widget that collects a name and a number
- Not a script that reads the FAQ back to people
- Not a black box: every action it takes is logged
Every channel you use. One brain behind all of them.
WhatsApp, Instagram, the chat box on your site and your phone line are four doors into the agent we built for you. One knowledge base, one set of tools, one set of rules. Change a price once and it changes everywhere. Add a fifth channel next year and it arrives already knowing everything.
Where Qatar actually talks. Text, voice notes, images and documents, on your business number.
Direct messages and story replies. Same thread, same agent, same knowledge.
Web chat
The box on your site. No separate build, no second knowledge base to sync.
Phone
It picks up on the first ring, at two in the morning, in Arabic or English. Voice is a channel, not a separate product.
So the phone and WhatsApp never give different answers.
A voice note at eleven at night. A confirmed booking before midnight.
One conversation, from the message arriving to the customer being done. Scroll it, and every step the agent takes before it answers is named as it happens.
The work happens between the two messages.
In Arabic, at an hour when nobody is at the desk. The named steps are the work it did before it replied.
Illustration. Not a real client conversation.
- 0:14
- Understood
مساء الخير، أبغى أحجز بكرة الصباح، ونفس الشخص اللي جانا المرة الماضية لو ممكن.
Good evening. Can I book for tomorrow morning, with the same person as last time?
- What it did
- Checked what was actually free
- Matched the person she asked for
- Wrote the booking into the system
تمّ الحجز بكرة الصباح، ومع نفس الشخص اللي طلبتيه. وصلك التأكيد الحين.
Booked for tomorrow morning, with the same person you asked for. Confirmation sent.
Nobody re-keys this in the morning. The booking is already in the system.
And at any point, one message from your staff takes the conversation, and the agent steps out of the way.
Five parts. You should know all of their names.
Nothing here is a black box you rent. Each part is something you can point at, change, or switch off.
The Agent
The part that talks. It carries your tone, your rules, and the limits on what it may promise a customer.
The Knowledge Base
Everything it is allowed to say: services, prices, policies, hours. You edit it, and the agent flags anything outside it.
The Tools
What it can do inside your systems: check availability, book, cancel, reschedule, look a customer up. Every action is separate and logged.
The Handover
The rule for when it stops: a staff reply, a customer asking for a person, or a limit you set.
The Record
Every conversation, every action, every handover, written down and searchable.
The Monitor is not on this list
A separate product. The Monitor reads your own team's conversations so you can see what was said and coach on it. It never replies to your customers. It works with or without an agent.
See the MonitorIt knows when to stop. That is the part most people skip.
An agent that never hands over is a liability, not a saving. Ours watches the line in both directions. The moment one of your staff replies in a thread, it logs the takeover and goes quiet. No double answer, no cheerful follow-up on top of your manager's apology.
It comes back to the agent only after a long idle gap, and only if you configure it to.
When a human takes over
One staff reply flips that conversation to human mode. The agent stops sending and keeps recording.
What it refuses to do
It will not invent a price or promise a slot it has not checked. It says it will check, and escalates.
When it escalates
Complaints, refunds, anything angry, anything outside its tools. It routes to your team with the conversation attached.
Named clients, and what actually got built.
Three engagements running in production today, each one named with the client's agreement. What each system does is described here as a customer of theirs would experience it.
Dania Maids
Home cleaning services, Doha
Booking agent, operations dashboard and conversation monitor
A customer sends a voice note in Arabic at eleven at night. The agent transcribes it, checks which cleaners are genuinely free, matches the one that customer asked for by name, books the slot and confirms it on WhatsApp. Behind that, a dashboard the office runs the day on, and a monitor that reads every conversation the human team has.
Aram Dental Clinic
Dental clinic, Doha
Voice and WhatsApp appointment system with a waitlist
Patients book, cancel and reschedule by phone or by message, in Arabic or English. When a slot frees up the waitlist is offered it automatically. Conversations are scored for frustration as they happen, so the clinic sees the unhappy patient while they are still on the line rather than in a review a week later.
Shop Five
Retail and e-commerce
WhatsApp sales agent and an owner's voice assistant
On one side, customers search the catalogue, build a cart and check out inside WhatsApp. On the other, the owner asks out loud how yesterday went and gets the answer from the live database rather than from a report someone exported last week. The read side and the sell side are the same knowledge, asked two different ways.
Live in 2 to 4 weeks typically.
That is a first agent. A larger build takes longer, and we say so before you sign anything.
What we do
Map the conversations you actually get. Write the knowledge base. Connect the agent to the systems holding your bookings, customers and prices. Test it on real past conversations, then run it on a test number.
What you supply
Your price list and your policies. Access to the system that holds your bookings or orders. One person who can answer questions in week one. And a decision on what the agent may not do.
And then we stay on it
The weeks after go-live are when any agent is wrong in ways nobody predicted. So we read the conversations, close the gaps, and keep tuning. An agent nobody watches gets worse, not better.
The six questions everybody asks in the first meeting.
What is the difference between an AI agent and a WhatsApp auto-reply?
An auto-reply sends the same message to everyone and stops. An agent reads what was actually asked, looks it up in your knowledge base, calls your systems to find out what is true right now, takes the action the customer wanted, and records it. The test is simple. At the end of an auto-reply, the customer is still waiting for a human. At the end of an agent conversation, the booking exists.
Which languages and dialects does it handle?
Arabic and English by default, including switching between them mid-conversation, which is how people in Qatar actually write. On the phone, male and female voices in both languages. Urdu, Hindi and Tagalog are available on request, and are common for clients whose customers and staff speak neither Arabic nor English.
On the input side we handle Modern Standard Arabic, Egyptian, Levantine and Khaleeji, and the agent replies in Modern Standard Arabic or in a dialect configured to match your brand voice.
What happens when someone on my team steps in?
The agent stops. It watches outgoing messages as well as incoming ones, so the moment a staff member replies in that thread the conversation flips to human mode and the takeover is logged. The agent keeps recording but sends nothing further. It returns only after a long idle gap, and only if you configured it to return at all.
Where does my data live, and who can see it?
Your data is yours. We sign an NDA and a data processing agreement before anything starts, and we deploy either inside your own cloud account or in a dedicated tenant we run for you, never in a pool shared with another client. Conversations are encrypted in transit and at rest, and no client data goes into shared training sets.
We are built to enterprise practices, and we are in the process of certification. If procurement needs a named certificate written into the contract, raise it before you sign and we will tell you plainly where we stand.
What does it cost?
A fixed fee, scoped to what you want the agent to do, plus a separate monthly figure for running and tuning it. No per-seat licence and no per-message charge from us. What moves the number is how many of your systems the agent has to talk to and how many actions it is allowed to take, not how many messages you receive.
We quote against your actual scope rather than a bracket, so there is no price list on this page.
How long until it is live?
Two to four weeks for a first agent, typically. A larger project takes longer.
Week one is discovery and scope: your services, your recurring questions, and the systems the agent must talk to. Week two is build and integration: the knowledge base is written, the tools are connected, and the agent is tested on real past conversations. Weeks three and four cover WhatsApp template approval where it applies, a staged run on a test number, and a controlled launch with a human watching. Narrow use cases go faster.
Tell us the job,
and we will tell you what it takes.
Every agent we build starts with one question: what is the job you want finished without a person having to do it. Take thirty minutes. No deck, no pitch: we open a live system, and we answer whatever you ask.