How a conversational assistant can help your customers decide

Your customer lands on the product page with a specific question. If the answer isn't there, they go looking for it elsewhere. In this article we go through what a conversational assistant does, what the data says about its impact and what to look at before you get one.

In a physical store the customer asks and gets an answer. Whether that brand runs large, whether the fabric has any stretch, whether there's more of their size in the stockroom.

Online the route is different. They open the size guide, scroll down to the reviews, message support and get told they'll hear back within 24 hours. They end up outside the product page looking for an answer that should have been on it, and that's where the frustration starts. Not because they don't like the product, but because there's a doubt left hanging just as they were deciding.

Your customer already uses AI to shop, but not to decide

AI has entered the buying process faster than almost anyone expected. According to McKinsey and EuroCommerce's 2026 report Rewiring retail in Europe: The AI imperative, 61% of European consumers already use it to discover or evaluate products before buying them. In luxury the figure jumps: the latest Bain and Comité Colbert report puts at 82% the highest-spending customers who used AI on their last purchase, with 68% saying it helps them decide faster.

That said, using it isn't the same as handing over the decision. McKinsey asked 749 consumers across France, Germany and the UK how they feel about AI when they shop, and 56% said they're happy for it to suggest options, but they want the final say. And 47% don't want it placing a repeat order on its own.

So the question isn't whether your customer uses AI, but where they use it and what for. If they resolve their doubts on a general-purpose platform, that platform doesn't know your catalogue, doesn't know how your brand sizes and has no way of knowing what experience you want to give them.

And that also shapes how an assistant should behave inside a store. Its job is to answer the question the customer has in that moment, not to take the chance to sell them something else or decide for them.

What questions it actually answers

In practice, most of the questions that come up before a purchase fall into four categories:

Size and fit: How the brand runs, whether to size up or down, how that cut sits, which size to pick.

Product: What it's made of, whether it shrinks, how to wash it, whether the colour in the photo matches the real thing.

Delivery and timing: When it arrives, whether it'll get there before a specific date, whether it can be collected from a pickup point.

Returns: What happens if it doesn't fit, how long they have to send it back, whether the exchange is free.

All four have something in common: these are questions the customer asks before paying, not after. And when they don't find the answer, the purchase gets postponed or ends up somewhere else.

What brands already using one are seeing

Gorgias's State of Conversational Commerce 2026 report gathers responses from 400 ecommerce decision-makers and data from more than 16,000 brands.

79% of brands say AI-driven conversational commerce has increased their sales. And 84% say the strategic importance of this channel is higher now than it was a year ago.

But there's one detail in that report worth pulling out. When asked to rank where the biggest return came from, improved conversion wasn't the first answer:

• 38% pointed to customer service efficiency
• 23% to retention and loyalty
• 20% to improved purchase rates

In other words, brands aren't seeing the value of conversational commerce purely in selling more. A significant part of the impact is that the team stops answering the same questions over and over.

Most support conversations in ecommerce revolve around doubts about product, delivery, sizing or returns. If the assistant can resolve them on the product page, the team gets time back for the conversations that genuinely need a person.

What the figures say and what you should measure

If you look up how much conversion rises with a conversational assistant, you'll find all sorts.

A lot of the problem is in how it's measured. The most common approach is to compare people who use the assistant with people who don't, and the result comes out inflated, because anyone who opens a chat already had more purchase intent. And nearly all those figures are published by whoever is selling the technology.

Rather than looking for a benchmark percentage, it's more useful to measure it in your own store. These are the metrics that help:

Conversion with the assistant versus without it: Compare users who see the assistant with a similar group who don't, over the same period. It's the only way to know whether the effect comes from the assistant or from the type of customer who opens it.

Open rate: How many people who reach the product page actually interact with the assistant. If it's very low, the problem may not be the assistant but how and where it appears.

Questions per conversation: It tells you whether the customer is using the assistant for a one-off query or holding a conversation to resolve several doubts before buying.

Product-related support tickets: This is one of the metrics that can move soonest. If the assistant answers product questions well, some of those queries should stop reaching your team.

Size-related returns: Over the medium term, it can tell you whether the conversations about sizing are genuinely helping customers choose better.

On their own, none of these metrics says much. Together they give you a fairly clear picture of whether the assistant is adding anything.

What to know before you implement one

Where it appears: The doubt almost always comes up on the product page, looking at photos and hunting for the size. If the assistant sits somewhere else on the site, or the customer has to click several times to reach it, chances are they won't use it. And if they don't use it, it hardly matters how well it answers.

How long it takes to respond: A wait of several seconds in the middle of a purchase decision can create more friction than the assistant removes. The first tools of this kind took a while to respond, and that's where they lost most of their usefulness.

What it knows: This is where you see whether the assistant is genuinely integrated. One that answers with generic information saves the customer nothing, because that information is already on the page or a search away. The difference shows when it knows the stock available, the specific sizing of your brand and the real return conditions.

What it knows about the person asking: Those conditions don't have to be the same for every customer. Asos made that visible in early 2026: every customer sees their own return rate when they log in and when they place an order, worked out on the value returned over the previous twelve months rather than the number of parcels (Retail Week, 2026). Asos uses it to charge the highest returners, but what matters is what it proves: that information can be personal, and it can appear before payment. An assistant connected to that logic answers what happens if it doesn't fit for the person actually asking, instead of linking to the general policy.

Whether it pushes or answers: You've already seen that 56% want the final say. An assistant that pushes towards a purchase instead of resolving the doubt loses exactly the trust it was building.

Does a conversational assistant reduce returns?

It can reduce them if it resolves the sizing question before payment, and it won't touch them if it only answers generic doubts. The difference is whether the assistant knows your brand's actual sizing or just repeats what's already on the page.

Sizing is the leading reason for returns in online fashion. According to our State of Returns 2025 report, 34.63% of returns happen because the garment doesn't fit, and another 21% because the customer ordered several sizes at once.

That's where an assistant can do something a size guide can't. A guide is one table for the whole catalogue. An integrated assistant answers about that specific garment, with its cut and with what other customers have done with it.

What to expect and what not to

Don't expect it to move in a fortnight. Size-related returns are read by season, not by sprint. What does move sooner are the support tickets about sizing, which is the signal that the assistant is answering that question instead of your team.

From answering doubts to guiding the purchase

Tailor is our conversational assistant and it works inside the product page.

It answers questions about sizing, materials, delivery and returns using your store's real information, not generic data. Because it's integrated into your ecommerce, it can take into account the product the customer is looking at, the stock available and other store data. It can also add the product to the basket within the same conversation.

And some questions can't be answered with text alone. That's why Tailor also lets the customer see how the garment looks on them, in photo or on video.

The impact on conversion depends on the category, the product and how it's implemented, so we're not going to give you a percentage. What is clear is that the doubt exists, that the customer is already looking for the answer, and that right now they're finding it outside your store.

Frequently asked questions

What is a conversational assistant in ecommerce?

It's a tool that answers in real time the questions a customer has before buying, within the shopping experience itself. Unlike a chatbot with predefined answers, it understands natural language and can work with the store's real information: products, stock, sizing, delivery times and return conditions.

Do chatbots actually work in an online store?

A chatbot with closed answers, or one that simply redirects the customer, can create more friction than it removes. An assistant that knows the catalogue and resolves the doubt on the spot can stop the customer having to leave the page to find the answer.

Is a store's own assistant better than a generic one?

An integrated assistant can work with information specific to your store, such as the catalogue, stock, sizing or returns policy. A generic one can only answer with what it finds published, so its response rarely adds anything the customer couldn't have found on their own.

How much does an AI assistant increase conversion?

There's no universal figure. The available studies use different methodologies and many don't have control groups. Results also depend on factors like category, product, traffic and implementation. That's why it's more useful to look at the impact within each ecommerce's own context.

What questions does a conversational assistant answer?

Mostly questions about the product and the buying decision: size and fit, materials and features, availability, delivery times and return conditions. These are doubts that can come up right before payment and that, if they go unanswered, can lead the customer to postpone the purchase or look for the information somewhere else.