Recommending products on WhatsApp to undecided shoppers

An undecided shopper messages you because the product page didn't answer one specific doubt: will it fit, does it work for my case, which of these two should I pick. It isn't a lack of interest; it's a lack of a tailored answer.

In a physical store, the associate would ask two questions and pull a couple of items off the shelf. Online, the shopper has forty tabs open and a chat that answers tomorrow. Every hour of waiting cools the purchase, as we cover in why first response time matters.

Before recommending anything, ask one to three questions that narrow the catalog: who it's for, what they'll use it for and roughly what they want to spend. More than three feels like a form, and people leave.

Example: a skincare store gets "I want something for dry skin." A weak reply pastes the whole category link. A good question is "Is it for face or body, and do you want something light or richer?" With that answer you can name a specific product.

A practical rule: every question must change the recommendation. If the answer wouldn't change what you offer, don't ask it.

Show two or three options, never the whole catalog, and say in one sentence why each one fits. Indecision feeds on abundance; a short list with a reason shrinks it.

It works well to give one main pick ("the closest match to what you told me") and one alternative that differs in price or style. Each gets a name, a reason and a link. Skip empty adjectives like "great quality": the shopper needs a verifiable reason, such as material, size or the use it was designed for.

A useful recommendation can only come from real data: what's in stock, in which size or variant, at today's price. Recommending a sold-out product is worse than recommending nothing, because it turns doubt into frustration.

That's why the agent needs to read the store's catalog rather than rely on a list someone updated last month. At heysteff, Steff works with the Shopify integration (catalog, orders and carts) and the brand's knowledge base, so it answers with what the store has today. To see how this kind of assistant is built, visit AI shopping assistant.

A person should step in when the purchase is high-value, when the question touches health, safety or warranty, or when the shopper asks to talk to someone. AI handles the normal range well; human judgment matters more at the edges.

Two mechanisms help: automatic escalation to a human based on keywords and rules, and copilot mode, where the AI drafts the suggestion and a person decides whether to send it. For stores with delicate products, starting in copilot mode is a prudent way to build trust before granting more autonomy.

The honest measure is how many recommendation conversations end in a purchase, not how many messages were sent. Track three things: how many shoppers click the recommended product, how many buy afterwards and which questions keep repeating.

Those repeated questions are gold: if ten people ask the same thing, the product page is missing that information. And if the shopper leaves without buying, a follow-up reminder can help; that's what abandoned cart recovery over WhatsApp is for. You can also read how to handle pre-purchase questions in this article on pre-purchase questions.

The most common mistakes are replying with a generic link, recommending without asking anything, showing too many options and failing to mention low stock or an unavailable variant. They share one cause: treating the shopper as just another inquiry instead of a person about to decide.

Tone is another one. A message written like a brochure ("discover our amazing collection") sounds like advertising; one that picks up what the shopper said ("you mentioned travel, so I'd suggest the lightweight model") sounds like service. An AI agent's brand tone can be trained to keep that closeness without losing the store's voice.

Finally, there's the mistake of not closing the loop: recommend, get an "I'll think about it" and never write again. A short follow-up the next day, only when the shopper started the conversation, recovers sales that would otherwise go cold.

A well-handled conversation usually has four steps: a short greeting, one or two questions, two or three options with reasons and a clear invitation to buy or keep talking. All in a few messages, without making the shopper repeat themselves.

Picture a running-shoe store. The shopper writes "I don't know which ones to pick for running." The agent asks whether they run on roads or trails and how many miles a week. With that, it suggests a main model and a cheaper alternative, says which sizes are available and leaves the link. If the shopper mentions an injury, the agent doesn't improvise advice: it hands off to someone on the team.

That last decision, knowing what not to answer, is what separates a useful assistant from a risky one, and it's set before the agent goes live, not after.

For AI to recommend well, each product needs data that answers the typical doubts: sizes, materials, who it's for, what it's good at and how it differs from a similar item. If the page only says "beautiful and high quality," the agent has nothing to argue with.

Spend an afternoon on the ten products that generate the most questions. Add a clear size chart, a one-line "best for" and the questions you hear most. That information serves your store page, search engines and the agent's knowledge base at once, so the effort pays off three times.

Can you recommend products on WhatsApp without it feeling like spam? Yes, if you're replying to a question the shopper started and you offer a few options with a clear reason. Spam is sending lists nobody asked for.

Does an AI shopping assistant replace the sales associate? Not entirely. It handles frequent questions and mid-priced purchases well; complex or high-value sales are better left to a person, with the AI as support.

What do I need to get started? A tidy catalog, product descriptions with useful details (sizes, materials, uses) and clear rules for when to hand off to a person. With that, the agent can already guide an undecided shopper.

• AI shopping assistant
• AI agent for Shopify
• AI agent for ecommerce
• Answering pre-purchase questions