WhatsApp bot or AI chatbot? Why ours deliberately isn't an AI

For questions about a customer's own order, invoice or booking, use a rule-based WhatsApp bot that reads your records: it gives the right answer every time and can't make one up. Use AI only for open-ended questions (products, policies, "do you do this?"), and only from content you control, with a clear hand-off to a person. Most UAE businesses need the first, and many are sold the second.

The WhatsApp bot we built into SARZ Command is deliberately not a language model. Here's why, and when we'd choose AI instead.

What the customer is really asking

Look at what customers actually message a business about. Most of it falls into a few questions:

• Where's my order or shipment?
• What do I owe? Can you send the invoice?
• Can you send my documents?
• I need a quote.
• Can I talk to someone?

Each one has a single correct answer that already exists in your system. A bot doesn't need to be creative to answer it. It needs to look it up and reply in under a second.

How the SARZ Command bot works

The bot recognises six intents: menu and greeting, track a shipment, my invoices, my documents, get a quote, and talk to a person. It identifies the client by the number they're writing from, so it only ever returns that client's own shipments, invoices and documents. Clients never log in.

Because it isn't a language model, a client asking where their cargo is gets the real stage and ETA every time, with no chance of an invented tracking number. A quote request becomes an enquiry in the sales queue, and "talk to a person" hands off to the assigned rep. You can see it in the cover image: a real conversation from the system, with sample data.

Where AI goes wrong in customer service

Language models are very good at sounding right. That's the problem when the answer has to be right. An AI bot asked "when will my order arrive?" can produce a confident date that isn't in your system. Asked about a refund policy, it can describe one you don't have. Each of those is a customer who now distrusts you, or a promise your team has to honour.

When AI is worth it

AI earns its place when the questions are open-ended and the answers don't depend on the customer's account:

• Product discovery: "which one is better for a small office?"
• Long FAQs: delivery areas, warranties, how a service works.
• Understanding messy messages and routing them to the right flow or person.

Even then, the AI should answer only from content you've approved (your FAQ, product pages, policies), say when it doesn't know, and hand off to a person easily.

The best of both

The setup we recommend for most UAE businesses is a hybrid:

1. Rules for anything about the customer's account: orders, invoices, bookings, documents.
2. AI, optionally, for general questions, answering from approved content only.
3. A person for everything else, with the conversation history attached so the customer never repeats themselves.

What it costs to run

Every WhatsApp bot pays Meta's message pricing for templates (see our WhatsApp Business API guide). An AI bot adds AI usage on top: Meta prices its own Business Agent messages by tokens, and third-party AI is billed separately (Meta pricing). A rule-based bot adds no AI cost at all.

Want a WhatsApp bot that answers from your own data? See WhatsApp chatbots, or book a SARZ Command demo to see this one working.