Answering pre-purchase questions: the sale that gets lost in the inbox

Because they are the highest-intent messages an online store receives. A pre-purchase question is any inquiry made before paying: is my size in stock, how long does shipping take, will this work for my case. The shopper has already picked the product; one answer stands between them and the checkout button.

Unlike a complaint, there is nothing to repair: there is a sale on pause. Every hour without a reply raises the odds that the person opens another tab and compares. That is why these messages deserve their own queue and priority, separate from post-purchase issues and complaints.

In almost any store the doubts fall into five groups: availability (size, color, stock), shipping (timing, cost, coverage), product (materials, compatibility, use), payment and guarantees (installments, exchanges, returns) and trust (is the store real, will the order arrive). All five share one thing: the answer already exists somewhere, whether in the catalog, in the shipping policy or in the head of whoever handles the inbox.

A useful exercise is to open your inbox and tag the last fifty messages sent before a purchase. Usually three or four questions account for most of them. That list is the starting point for everything else: your FAQ page, your saved replies and what you teach an AI agent.

Pay attention to how the questions are phrased, too. Shoppers rarely write “what is your return policy”; they write “can I send it back if it doesn't fit?”. Keeping their own wording in your saved answers makes replies feel human and helps any assistant match the question to the right fact.

An FAQ page covers generic questions, but pre-purchase doubts are usually specific: “will the medium jacket fit if I'm 6 ft tall?” or “will it arrive in Denver by Friday?”. Nobody finds that on an FAQ page, and shoppers who don't find it either message you or leave.

The page is still worth having, and it helps search engines and AI assistants understand your store, but it works better as a source of truth than as a sales channel. What converts is the answer reaching the person, in the channel where they are already talking to you, with the right fact for that product.

Anything written down and verifiable can be answered automatically: stock, variants, price, shipping times, exchange policies and product specs. An AI shopping assistant connected to your Shopify catalog checks those facts at the moment the question arrives, instead of repeating an answer that went stale.

What shouldn't run unsupervised is anything that depends on judgment: a price exception, sensitive medical or technical advice, a special order or an upset customer. There, the AI's job is to prepare the context and hand the conversation to a person. To define when, see how to set human escalation rules, and for the bigger picture read what to automate and what to keep human.

It matters more than for any other kind of inquiry, because a purchase decision has an expiry date: while the shopper is looking at the product, your reply competes only with their distraction; an hour later it competes with every other store. Measuring first response time on pre-sale messages alone tells you more than a general average. For more, see why first response time matters.

Speed also doesn't mean a generic reply. “Thanks for writing, we'll get back to you soon” resolves nothing. A good reply carries the requested fact and, when relevant, the next step: the link to the right variant or the one question needed to recommend well.

The first is replying with pasted text that misses the real doubt: the shopper asks about sizing and gets the opening hours. The second is not checking stock before promising, which turns a closed sale into a refund. The third is answering well but leaving the next step unclear, so the conversation dies at “thanks”.

There is a fourth, quieter mistake: not learning from the questions. If ten people ask the same thing, the problem probably sits on the product page, not in support. Every repeated question is a signal to improve the description, the size guide or the shipping page. Over time, good pre-sale support reduces how many doubts arrive at all.

Finally, watch the tone. A jewelry store and a sportswear store shouldn't sound alike, and a fast reply doesn't have to sound like a form. Write two or three model answers in your brand's voice and use them as the reference for every person or AI that replies.

Picture a Shopify clothing store that gets this WhatsApp message on a Saturday night: “Does the black jacket come in L, and will it arrive in Austin before Thursday?”. Without automation, the reply comes on Monday and the shopper already bought elsewhere. With an agent connected to the catalog, the L variant's stock is checked, the reply uses the store's shipping policy, and the direct product link goes out with it.

If the shopper asks something the policy doesn't cover, such as a rush delivery outside coverage, the agent hands it to someone on the team with the full conversation. The customer doesn't repeat themselves, and the team only steps in where their judgment adds value.

Steff answers these questions on WhatsApp, Instagram, Messenger and Gmail from a single inbox. With the Shopify integration it reads catalog, orders and carts and replies with your store's information. Whatever it doesn't know comes from your knowledge base, where you load policies, size guides and approved answers, and where you set the brand tone. With copilot mode you can start by letting the AI suggest while a person approves each reply.

A low-risk way to begin: pick the five questions that repeat most, load verified answers for them, and spend a week reviewing which conversations were handed off and why. That log tells you what's missing from the knowledge base better than any guess.

What is a pre-purchase question? It is any inquiry a shopper makes before paying: size, stock, shipping, warranty or compatibility. These messages tend to carry more buying intent than others because the product has already been chosen.

Can an AI answer product questions without getting them wrong? It can, when it answers only from verifiable catalog data and store policies and hands off to a person when a reply needs judgment. Starting in copilot mode lowers the risk while you check quality.

How many questions should I prepare to start? The five or ten that repeat most in your inbox are enough. Start there, review the handed-off conversations and expand the knowledge base based on what is missing.

• AI shopping assistant
• AI agent for Shopify
• Knowledge base
• First response time