WhatsApp AI for Shopify Stores in Lebanon and the Gulf

A Shopify store in Beirut, Dubai or Riyadh usually has a checkout that works perfectly and a WhatsApp number doing most of the actual selling. That is not a failure of the store. It is how the region buys, and the stores that win are the ones that stop fighting it.

WhatsApp AI for a MENA Shopify store means connecting a language model to your Shopify catalogue, inventory and orders, so it can answer product questions, check real stock, recommend and sell, and return order status in Arabic, English or Arabizi, inside the 24 hour window where replies are free. It matters more here than in Europe or the US because Gulf and Levantine shoppers routinely start and finish a purchase in chat rather than at the checkout, and because cash on delivery keeps a human in the loop. The work is mostly not technical: it is cleaning the catalogue, writing down undocumented policies, and building a clean handover to a person. Expect two to four weeks.

Why the checkout is not where the sale happens

In most of the region, a product page is the start of a conversation, not the end of one. The pattern is consistent enough to plan around:

• Trust is transacted through people. A shopper wants confirmation from someone before paying, especially at higher basket values. Cash on delivery, still dominant across much of the Levant and parts of the Gulf, makes that instinct rational rather than sentimental.
• Sizing and fit are negotiated. "Is this true to size", "I am usually a 38 in Zara", "will this fit my mum". These are not FAQ entries. They are judgement calls, and they are where the sale is won.
• The phone is the storefront. Instagram and TikTok drive discovery, WhatsApp closes. The website is often just where the product photos live.
• Delivery is the real question. Not whether you deliver, but when, to which area, and what happens if nobody is home.

None of this is a problem to be solved. It is a channel that converts better than your checkout, staffed by people who cannot answer at midnight.

What actually breaks

Stores that automate this badly break in the same four ways.

The catalogue lies

Your automation is only as truthful as the data behind it. If Shopify says a size is in stock because someone forgot to decrement it after a market stall weekend, the AI will confidently sell it. That is worse than silence, because now you are cancelling an order that a person promised.

Fix the catalogue first. Variant names, stock accuracy, and the difference between out of stock and discontinued.

Arabizi is treated as a typo

A large share of messages arrive in Latin-script Arabic. 3ande, shu ha, keefak, bade hayda bel aswad. The number substitutions are not noise, they are the writing system: 3 for ain, 7 for haa, 2 for hamza.

Keyword matching cannot survive this, because there is no single correct spelling to match against. A menu based bot handles it by refusing to, and pushing everyone to buttons. Language models handle it natively, which is the single strongest technical argument for that approach in this region specifically.

The handover is an afterthought

The moment a customer says something the automation cannot handle, three things must happen: a person is notified, that person can see the entire conversation, and the customer knows a person has arrived. Miss any of the three and you have built a very fast way to annoy people.

Nobody reads what came back

The least visible failure and the most expensive. A month of conversations tells you which products people ask about and do not buy, which sizes you are constantly out of, what they think you sell that you do not, and which objection kills the most baskets. Almost every store throws this away.

What a working setup looks like

What it is not right for

Worth saying plainly, because the wrong fit wastes everyone's time.

If you get a handful of messages a day, you do not need this. A fast human is better and free. If your catalogue is a mess and nobody has time to fix it, automation will amplify the mess rather than hide it. And if you want to configure everything yourself tonight, an off the shelf shared inbox from the comparison of tool categories will suit you better than a managed build.

Where to start this week

Export a week of your WhatsApp messages and sort them into three piles: questions with exactly one right answer, questions that need a lookup, and questions that need a person's judgement.

The first pile is what you automate first, and it is usually larger than anyone expects. The second is what makes the Shopify connection worth building. The third is what your team should have been spending its whole day on.

That sorted list is the specification, and it is the same thing we start from when we build what plum handles for a store. If you want to see how it behaves before committing to anything, message it and ask it something difficult.