Helping shoppers decide without automating the joy

Three years ago, I would have told you that nobody wanted to talk to a retailer. Now I talk to my machines all day.

That change was on my mind when I sat down with Microsoft's Ashima Chetan at the CommerceNext AI Forum. I really enjoyed the conversation and was grateful for the perspective she brought from helping shoppers across Microsoft's partner and retail channels.

What made the discussion useful was the space between what the technology can do and what customers actually want.

I see a growing appetite for conversational discovery. Being able to describe your situation in your own words can be enormously helpful. But that doesn't mean every shopper wants a conversation, every time. Some know exactly what they want. Others want a few guided choices. Others need to talk the decision through.

Ashima described a guided laptop experience where sensible questions still felt like a quiz, and shoppers left before reaching a recommendation. That is a useful challenge for anyone building these experiences. A question can be relevant and still be one question too many. The experience has to adapt to the person, including knowing when to get out of their way.

The next question is whether the customer trusts the guidance enough to act on it. A fluent answer isn't enough. They need confidence that the advice reflects their needs and that the product is right for them.

I used a fridge as an example. The dimensions tell you whether it fits your kitchen. Owner reviews can tell you whether its noise is noticeable in the space where you relax. Connecting those different kinds of knowledge to someone's open-plan home makes the advice useful. The assistant also needs to be clear about the evidence behind it.

This is why we've brought our approach together in TRACE: a Transparent, Relevant, Adaptive and Consistent Experience. It is a practical framework for earning trust, founded on dependable product knowledge. Explain the advice and its trade-offs. Keep it relevant to the decision. Adapt when the customer tells you something new. Carry context forward with permission, and keep the facts and reasoning consistent. Those qualities need to hold across the whole experience, from search and browsing to comparison and conversation.

We should apply the same honesty to measurement. Strong conversion among people who choose to use an assistant doesn't, by itself, tell us what the assistant added. Ashima's point about a proper holdout and measuring all eligible visitors matters. We need to understand who benefits, who encounters friction, and whether the overall experience improves.

That customer perspective becomes even more important as agents gain the ability to act on our behalf.

13 best ecommerce search engines in 2026 (Backed by data—and why most still fail your customers)

Read more

My CRO was looking forward to shopping in Milan. The anticipation, discovery and choosing were part of what made the trip appealing. Making the purchase easier is one thing. Removing the experience altogether could take away the very thing the customer was looking forward to.

We shouldn't automate the joy.

How much someone wants to delegate will depend on the person, the occasion and the category. Routine replenishment may be welcome territory for automation. Choosing something personal may be a different matter. There is still a great deal to learn about where people want help and where they want to remain involved.

For me, that brings the conversation back to the same place: understand the customer, make discovery more accurate and reliable, earn their trust, and let them choose how much to delegate.

Thanks again to Ashima and CommerceNext for a conversation that gave us plenty to think about.