WhatsApp AI customer service for UAE e-commerce: order status, COD confirmation and returns

TL;DR: For a UAE online store, WhatsApp is the customer service channel whether you planned it or not. An AI assistant earns its place when it is connected to your store platform and courier, not just your FAQ: it answers "where is my order?" from live data, confirms cash-on-delivery orders before they ship, starts returns within your policy, and answers product questions from the catalogue in Arabic and English. Refunds, complaints and exceptions go to a person with the conversation summarised. Since October 2026, design it to use as few messages as possible, because WhatsApp service replies now cost money past the monthly allowance.

Why WhatsApp is the e-commerce support channel in the UAE

UAE shoppers message stores on WhatsApp because that is where they message everyone. They ask before buying ("do you have this in a 42?"), while waiting ("where is my order?"), on delivery day ("the driver is calling, can he come at 6?") and after ("this doesn't fit, how do I exchange it?"). A store that answers in hours loses sales to one that answers in seconds, and a support team that answers every "where is my order?" by hand is spending its day reading tracking pages aloud.

Most of these questions have answers that already exist in your systems. The order status is in Shopify, WooCommerce, Salla or Zid. The delivery status is in the courier's system. The return policy is written down. The stock level is in the catalogue. An AI assistant connected to those systems can answer them instantly and correctly; one that is not connected can only apologise and promise someone will get back.

Order status: answering "where is my order?" from live data

"Where is my order?" is usually the most common question an online store receives. A connected assistant handles it in one exchange:

1. It identifies the customer by their WhatsApp number and finds their recent orders, or asks for the order number if the number does not match.
2. It reads the order status from the store and the shipment status from the courier's tracking feed.
3. It replies with a plain answer: dispatched yesterday with the courier, out for delivery today, or delayed with the reason the courier gave.
4. If the order is stuck, for example no courier scan for longer than your normal handover time, it flags the order to your team rather than reassuring the customer that all is well.

That last step is what makes it useful to operations, not just to customers. The assistant becomes an early warning for shipments that have gone quiet.

COD confirmation: stop shipping orders nobody wants

Cash on delivery is still common in the UAE, and with it comes a familiar cost: orders placed on impulse, with wrong addresses, or as duplicates, which go out with a courier and come back refused. Every one of those has a shipping cost both ways and ties up stock.

A confirmation step on WhatsApp catches many of them before dispatch. When a COD order is placed, the system sends an approved utility template with the order summary, the delivery address and three buttons: confirm, change, cancel. Confirmed orders are released to fulfilment. A change updates the address or delivery preference in the order. A cancellation stops the order before a courier is booked. No reply within your chosen window flags the order for a call or a hold, according to your rule.

The same conversation can collect what couriers in the UAE most often need and least often get: a precise location pin, a landmark, a building and flat number, and a preferred time. Fewer failed delivery attempts follow from better addresses, not from better drivers.

Returns and exchanges within your policy

Returns are where a good assistant saves the most back-and-forth. It checks the order date against your return window, checks the product against any exclusions (final sale, hygiene items, customised products), asks for the reason and photos where your policy requires them, and offers the options your policy allows: exchange for a size, store credit, or a refund request.

It then creates the return in your store or helpdesk and tells the customer the next step, whether that is a courier pickup or a drop-off point. The refund itself is the action to gate. Either a person approves it, or you agree a narrow automatic rule, such as an unopened item inside the window, and the assistant applies only that rule. Everything outside it goes to your team with the photos and the summary attached. Our guide to human-in-the-loop design explains how to draw those lines.

Product, size and stock questions from the catalogue

Pre-sale questions are revenue, and they arrive at all hours. An assistant grounded in your catalogue can answer whether a product is in stock in a size or colour, what it is made of, how the sizing runs, how long delivery takes to a given emirate, and which payment options are available, including instalment options such as Tabby or Tamara if you offer them. It can send the product link or a payment link so the customer completes the order without leaving the chat.

The important word is grounded. The assistant answers from the product data and your written policies, not from general knowledge, and when the answer is not there it says so and passes the question on. An assistant that guesses a size or invents a delivery promise creates the returns it was meant to prevent. We explain how grounding works in what RAG is and why it matters.

Arabic and English, properly

UAE customers write in English, Modern Standard Arabic, Gulf dialect, Arabizi and mixtures of all four, sometimes in the same message. The assistant should reply in the language the customer used, understand dialect questions like "وين طلبي؟" as readily as "where's my order", and use product names and policy text that were written in Arabic rather than machine-translated. If your catalogue is English-only, decide how product names should appear in Arabic replies before launch, not after the first confused customer. The engineering behind this is in why Arabic breaks most chatbots.

What should go to a person

• Refunds outside an agreed automatic rule.
• Complaints and any conversation where the customer is upset. The assistant should recognise frustration and hand over early, not late.
• Damaged or wrong items, after collecting photos and the order details.
• Exceptions to policy, such as a return outside the window for a loyal customer. That is a business decision.
• Payment disputes and anything involving card details. The assistant should never ask for a card number in chat.

The handoff lands in a shared inbox where your team sees the whole conversation and a summary, takes over in the same WhatsApp thread, and can hand back when done.

The integrations that make it work

System	What the assistant reads or writes
Store platform (Shopify, WooCommerce, Salla, Zid, Magento)	Orders, customers, products, stock; order notes, tags and status changes
Courier (direct API or aggregator)	Tracking events, delivery attempts, pickup bookings for returns
Helpdesk or shared inbox	Handoffs, conversation history, return and complaint tickets
Payment provider	Payment links and payment status, never card details
WhatsApp Cloud API	Messages, interactive buttons and lists, approved templates

The integration work is most of the build. It is also what you should ask any vendor about first; an assistant that cannot read your orders is a FAQ page with a chat window.

Design for the new WhatsApp pricing

Since 1 October 2026, Meta charges for service replies beyond 1,000 per business number per month, and utility templates sent inside an open customer service window are no longer free. E-commerce stores feel this more than most, because so many conversations are inbound. Three habits keep it under control: answer in one complete message rather than several fragments; use buttons and lists so the customer picks rather than types back and forth; and combine order updates into fewer templates where you can. The details are in our post on the October 2026 WhatsApp pricing change.

Abandoned cart reminders are marketing templates, charged per message and requiring the customer's marketing opt-in. One well-timed reminder is reasonable; a sequence of them gets your number reported, which lowers its quality rating and can restrict what you are allowed to send.

How to measure whether it is working

Before launch, record a baseline for a few weeks: the share of support conversations that are order status questions, average first response time, the share of COD orders refused at the door, and the number of support hours spent per week. After launch, track the same figures, plus the share of conversations the assistant resolved without a handoff and a weekly sample of conversations read by a person. We cover the full set of metrics in how to measure whether your AI assistant is working.

How Soluvide builds this

We are an engineering studio in Abu Dhabi. For online stores we build WhatsApp assistants on the official Cloud API, connected to the store platform, the courier and the team's inbox, in Arabic and English. Refunds and exceptions are gated behind human approval, every action is logged, and the WhatsApp Business Account, code and data stay in the store's name. See our e-commerce page and our WhatsApp AI assistant for how the pieces fit, or message us on WhatsApp with the question your team answers most.