Shoppers Who Use a Store's Own AI Assistant Spend 8% More. Accurate Answers Keep Them

Customer ExperienceBlog
Shoppers Who Use a Store's Own AI Assistant Spend 8% More. Accurate Answers Keep Them

On 12 November Salesforce published a holiday survey of more than 2,500 shoppers in the US, the UK, Canada and Australia. It found that 39% of shoppers have used a brand’s own shopping assistant on an ecommerce site, and 80% of them had a good experience. Shoppers who engaged with one placed orders 8% larger on average and bought 23% more items than the average order.

The same report carries the line that matters most for a small store: most shoppers will abandon an AI assistant after just two or three mistakes. Salesforce sells these agents, so read its numbers as a direction, not a forecast for your shop. The direction is still clear. The assistant earns the bigger basket only while its answers are right.

The gain comes from answers, not from having a bot

In a store, a shopper asks whether a jacket comes in a larger size, whether it will arrive before the 24th, whether it can go back to the shop. Online, those questions used to end in a contact form or an abandoned cart. An assistant that answers them in the moment keeps the shopper moving, and Salesforce found that 75% of shoppers with a good experience felt more loyal to the brand afterward.

More visitors now arrive with those questions ready. Salesforce reports traffic from AI tools to ecommerce sites up 104% year over year, and shoppers referred by AI more than twice as likely to convert as typical ones. They researched somewhere else and came to your site to confirm one detail.

Where the mistakes come from

The errors retailers report are rarely exotic. In a Talkdesk survey of 312 US retail decision-makers, commissioned by Talkdesk and run by Centiment in September, 58% said their AI had made errors, and 29% of those lost a customer because of it. A quarter cited inaccurate product recommendations, and 31% said their AI had taken an incorrect action on a customer account, such as an improper refund.

Two patterns sit behind most of that. The assistant answers from stale or incomplete information: a return window that changed, a size that sold out last week. Or it is allowed to do things that should stay with a person, like issuing refunds.

What to check before Black Friday

Write down the 20 questions your staff answered most last December: fit and sizing, stock, holiday shipping cutoffs, returns and exchanges, gift cards, store hours. Ask each one to your assistant, or to whatever answers your site today, and mark every answer that is wrong, vague or out of date.

Fix the source, not the reply. Put shipping cutoffs, the return window and gift card terms in plain text on your own pages, and make sure prices and stock come from your catalog, not from a document someone pasted in October.

Decide what goes to a person. Refunds, damaged orders and complaints should reach your team with a summary of the conversation. An assistant that only reads your catalog and policies cannot take a wrong action on an account.

Then keep measuring through the peak. Read a sample of conversations each week and count the wrong answers. Most shoppers give up after two or three, so every one you catch in a test is one a customer never sees.

An agent that answers from your own catalog and policies, on website chat and by phone, is what TalkRev builds. You can see how it works for retail stores.

Working in retail & e-commerce? See how TalkRev works for Retail & E-commerce.

One short email per new post

No ads. Unsubscribe anytime.

Start saving hours every
week with TalkRev

Reduce repetitive questions, capture more bookings, and deliver a faster customer experience — without adding headcount.