Your Customer Is Asking on WhatsApp Right Now. Who Is Answering?

Your Customer Is Asking on WhatsApp Right Now. Who Is Answering?

This series is about the tool every business owner is talking about and few understand in detail: the AI assistant that replies to customers. Seven parts unpack how it actually works, where it learns your company from, how you stop it from saying what should not be said, where your employee's role begins, what the Personal Data Protection Law requires, and how to measure its return in numbers.

But before the technology, the question that precedes it: what happens inside your company between the moment your customer's message arrives and the moment they read your reply?

WhatsApp is not a contact channel, it is your storefront

In the Saudi market, WhatsApp is the app a customer opens to ask about a price, book an appointment, follow up an order, and complain. Calls have become the exception, email is for formal correspondence, and few people fill in the form on your website. Everything else lands on one number in somebody's phone.

The consequence is that the first impression of your company is no longer your branch or your website — it is the speed and precision of a reply to a message. A customer who messaged three companies is not comparing service quality; they have not experienced it yet. They are comparing the one thing they can measure immediately: who answered, when, and how clearly.

And this is exactly where the gap shows. Most small and medium companies run this storefront with no system at all: a number on an employee's phone, conversations the manager never sees, messages that arrive after hours and wait for morning, and an employee who goes on leave carrying an entire customer history with them.

What a slow reply actually costs you

The cost of delay appears in no report, because it produces silence rather than complaints. The customer does not write to tell you they waited, found nothing, and went elsewhere — they simply disappear. Which is why so many owners underestimate it.

What research has established for years is that the value of an inbound lead decays fast. In a study published by Harvard Business Review covering more than a thousand companies, firms that responded to an inbound lead within an hour were far more likely to reach and qualify that customer than those who responded later. The logic is plain: a late reply arrives after the customer has already asked someone else and decided.

Then add what you lose inside the company itself:

• Your team's time. A large share of messages are repetitive questions: where are you located, how much is it, what are your hours, has my order shipped? An employee answering those dozens of times a day is an employee who is neither selling nor delivering.
• Peaks and evenings. Messages do not arrive evenly. They come in a wave after a campaign, and another in the evening once people finish their own working day — and you are usually closed for both.
• Inconsistency. Three employees give three different answers to the same question, and one of them grants a discount nobody approved.
• What vanishes without a trace. Conversations recorded in no system, so you cannot say how many customers reached you this month, let alone how many bought.

Why hiring does not solve it

The obvious fix is another customer service employee. It is partly right, but it does not address the nature of the problem, because message volume is not constant. An employee covers eight hours of a day in which your customer is active for twenty-four. And the peak when you need them most is precisely the moment they are busy with another customer.

On top of that, the employee you trained on your policies and prices may leave after a year and take what they learned with them, and you start from zero. Hiring addresses capacity; it does not address availability, consistency, or keeping knowledge inside the company.

The result is that most companies live in an uncomfortable middle: losing customers to slowness while paying salaries to people who spend half their time on repetitive questions a system could answer.

The old solutions, and why people abandoned them

This is not the first attempt at the problem. The market went through two generations before the current one:

• The canned reply: one message to everyone — thank you for reaching out, we will get back to you. It answers nothing; at best it confirms the message arrived.
• The menu bot: type 1 for prices, 2 for appointments. It works while the customer follows the path you drew, and falls over at the first real question outside the list. This generation is the reason for the aversion to automated replies you still see today.

What both share is that neither understands anything; they match patterns. The customer works that out within two messages and then writes: I want to speak to a person.

What changed to make the conversation serious

What is new is that language models now understand Arabic as it is actually written: in dialect, abbreviated, with typos, and split across two consecutive messages that complete each other. It has become possible to give the system your company's own knowledge so it answers from that rather than from general information. And the cost of running it is now within reach of a small company, not a project for a large enterprise.

None of which means everything marketed as artificial intelligence is serious. The difference between an assistant worthy of replying in your company's name and one that embarrasses you in front of a customer lies in technical and managerial details — where it knows things from, what stops it from inventing, and when it hands the conversation to a human. The rest of this series is about exactly those.

The Origami view

This is the reason we built Mahir: an AI assistant on WhatsApp that replies to your customers in Arabic within seconds, on your own company number, with no new number and no extra app on your employee's phone. You link your number by scanning a code once, teach it about your company by giving it your website link so it reads and fills in your details, then add your services, prices and frequently asked questions with answers you approve.

More importantly, it does not start replying to your customers straight away. The default is draft mode: it writes the reply and waits for your employee's approval before sending, so you can judge its quality on real conversations before it ever speaks in your name. When you are satisfied, you switch on direct replies yourself. The trial is a full week with every feature, no credit card.

Conclusion

The message arriving on your company number right now is not an enquiry; it is a quick test in which the customer decides whether to deal with you or with someone else. And you do not fail that test because of the quality of your service — you fail it because of a gap in time that nobody sees. In part two we open the box and explain how an AI assistant actually works: how it understands your customer's message, where an answer specific to your company comes from, and why it sometimes gets things wrong.

Sources

• Communications, Space and Technology Commission (CST) — Saudi Internet Report, indicators on messaging and communication app usage.
• DataReportal — Digital 2026 Saudi Arabia, platform usage figures for the Kingdom.
• Harvard Business Review — The Short Life of Online Sales Leads, on the effect of response speed on qualifying leads.