Williams-Sonoma's Shopping Assistant Converts 3x. What a Small Shop Can Copy

Two of America’s larger specialty retailers reported quarterly results within a day of each other this week, and both spent part of the call on the same subject: the assistant that now greets a shopper before any employee does. The figures they attached to it are the first ones specific enough to be worth copying.

What Williams-Sonoma actually measured

On the Q2 call, Chief Technology and Digital Officer Sameer Hassan said engagement with Olive — the assistant live on the Williams Sonoma site since last year — is up 700%, with revenue up 620%, and that shoppers who engage with it convert at three times the rate of those who don’t. Otto, the version rolled out across the Pottery Barn brands this month, resolves over 70% of conversations without handing anyone to a person. CEO Laura Alber described both as recommendation tools, not support tools.

The ordering of those two numbers matters. Seventy percent deflection is the metric chatbot vendors have sold for a decade, and it only ever describes cost. Three times conversion is a different claim: the assistant is not saving payroll, it is closing sales that were leaking. Hassan added that a visit where the experience is personalised now generates roughly nine times the revenue of an average visit, against two times a year ago.

Best Buy is moving the counter off its own site

A day later, Best Buy said it had completed a commerce integration with OpenAI — customers can now discover products, get recommendations and complete purchases inside ChatGPT. Jason Bonfig, who becomes CEO on November 1, called it a step toward “emerging AI-enabled commerce experiences.” The same call introduced Ask Blue, an assistant that compares products, checks compatibility, answers routine support questions and passes the shopper to a live specialist when it cannot.

Read together, the two calls describe one shift. The conversation that used to happen with a floor associate now happens in a chat window — and increasingly not on the retailer’s own website at all.

Where operators should start

None of this needs their budget. The mechanic is what transfers.

Pull your last 200 inbound messages — web chat, DMs, voicemail, email — and sort them into three piles: where is my order, returns and policy, and which one should I buy. The first two are the highest-volume, lowest-value contacts in retail, and both are answerable from documents you already have. The third pile is where the 3x sits, and it is usually the pile a small shop answers slowest.

Then write out your five most common pre-purchase answers once: sizing and fit, compatibility, what is included, delivery time, return window. That is an afternoon of work and it is most of the knowledge a shopper needs to finish a basket.

Put those answers where the question is actually asked — on the product page, in the DM, on the phone at nine in the evening — instead of behind a contact form somebody reads the next morning. For a shop selling both online and on a high street, that means one assistant covering both, because shoppers cross between them mid-decision and expect the answer to hold.

Finally, measure the right number. Not tickets deflected, but the conversion rate of sessions where a shopper asked something against sessions where they did not. Count only deflection and you will report the 70% while missing the 3x entirely.

The cost of waiting is quiet, which is what makes it expensive. Every unanswered “will this fit” after closing is a basket nobody comes back to, and no ticket is ever raised for it. Covering chat and phone together, without adding headcount, is the version of this a shop your size can run — and unlike a 700% engagement chart, it starts working the first evening someone gets an answer.