WhatsApp AI chatbot for businesses in Morocco: Darija, French and Arabic

In Morocco, customers don't fill in contact forms. They send a WhatsApp message, often a voice note, often in Darija mixed with French: "salam, wach mazal kayn had l'produit f taille M?" Shops, clinics, agencies, schools and real estate offices all end up with one or two people answering the same questions all day.

A WhatsApp AI chatbot can take most of that load, if it's built for how Moroccans actually write. Several Moroccan products now sell exactly this, which tells you the demand is real. Here's how I approach it as an engineer, what to watch out for, and when a custom build makes sense.

What a WhatsApp assistant can handle

The good use cases are the repetitive ones:

Product and price questions, answered from your own catalogue, not from the model's imagination.
Order status and delivery: "fin wslat la commande dyali?" answered by looking up the order in your system.
Appointments and bookings for clinics, salons, schools and garages, written straight into your calendar.
Lead qualification for real estate, car dealers and B2B: collect the budget, city and timing, then pass a summary to a salesperson.
After-sales and FAQs: returns, opening hours, payment methods, cash on delivery.

What it should not do: negotiate prices, handle angry customers alone, or make promises your team can't keep. Those go to a person.

Darija AI: handling Darija, French, Arabic and voice notes

This is where most generic chatbots fail in Morocco. A real inbox contains:

Darija in Latin letters with numbers (3, 7, 9 for ع, ح, ق), which has no single standard spelling.
Darija in Arabic script, Modern Standard Arabic from more formal customers, and French, often all in the same conversation.
Voice notes, which many customers prefer to typing.

The assistant has to detect what it's reading, search your content in a way that copes with five spellings of the same word, and reply in a register that fits: simple Darija or French for a shop, more formal Arabic or French for a bank or a clinic. Voice notes are transcribed first, then handled like text. I go deeper into this in building an Arabic AI chatbot that understands dialects.

Use the official WhatsApp Business Platform

There are two ways to connect a bot to WhatsApp:

The official WhatsApp Business Platform (the API), through Meta directly or an approved provider. You get a verified business number, message templates and a supported way to automate.
Unofficial tools that automate a normal WhatsApp account by driving WhatsApp Web. They're cheap and quick to set up, but they aren't a supported way to automate WhatsApp, and you risk losing the number your customers know.

For a business, the official route is the only one I recommend. Meta charges for some message types, so check current pricing for Morocco before you plan your budget.

Meta's 2026 rule on AI chatbots

Since January 15, 2026, Meta's WhatsApp Business terms bar "AI providers" from using the platform to offer general-purpose assistants, which is why ChatGPT and Copilot left WhatsApp. Meta has said this does not affect businesses using AI to serve their own customers. In practice: keep the assistant focused on your business (products, orders, bookings, support), give it a clear way to reach a human, and don't turn it into an open "ask me anything" bot.

Handover to a human

The handover is what makes customers trust the bot. The assistant passes the conversation to a person, with a short summary, when:

it isn't sure of the answer,
the customer is upset or asks for a person,
money, a complaint or personal documents are involved.

Your team picks it up in a shared inbox, sees the whole conversation, and the customer never has to repeat themselves.

Data protection: Law 09-08 and the CNDP

A WhatsApp assistant processes names, phone numbers and sometimes addresses or ID numbers, so Morocco's Law 09-08 applies. The CNDP has said that AI processing of personal data falls under the law, and transfers of personal data to countries without adequate protection need CNDP authorisation. If your AI model runs in the US or Europe, that's a question to settle before launch, not after. Options include masking personal data before it reaches the model, using a provider with suitable guarantees, or running the model on your own server. I compare these in private AI and on-premise LLMs. This is not legal advice; check your own situation with the CNDP or a lawyer.

The same applies if you sell in the Gulf or Egypt: Saudi Arabia's PDPL, the UAE PDPL and Egypt's Law 151 all regulate where customer data goes.

Off-the-shelf product or custom build?

A ready-made product is often the right start for a small shop with simple FAQs. A custom build makes sense when:

the bot needs to read and write to your own systems (stock, orders, CRM, booking software),
you want control over the model, the data and where it runs,
your volume makes per-conversation pricing expensive,
you need behaviour no product offers, for example several branches, several languages per branch, or specific approval steps.

Either way, start with one flow, like order status or bookings, measure how many conversations it resolves without a person, and expand from there. More on that approach in what to automate first.

Working together

I'm Driss Briksine, a software and ML engineer based in Morocco. I'm a native Arabic speaker, I work in Darija, Arabic, French and English, and I build AI assistants and automation end to end, from the model to the production software around it. See my work.

If your team spends its day answering WhatsApp, tell me about your inbox and I'll tell you honestly what an assistant could take off their hands.