I spend most of my week talking to teams putting AI to work inside big companies, plenty of them retailers. What strikes me is how fast they can change the customer experience now, and how little they’ve learned about the person they’re changing it for. So they keep building releases for a shopper somebody pictured a year ago. It’s like refitting the store overnight, every night, from last summer’s foot traffic map.
That old picture has a cost, and I hear it in the specifics they describe: a checkout redesign that loses conversion for reasons nobody can name, site search coming back empty for the words shoppers actually use, support tickets about a flow that works exactly as designed, an app rating sliding with no explanation attached. By the time any of that moves a number someone watches, the next release has already shipped. They file the drop under UI, performance or traffic quality. The cause sat further back. The shopper changed, and nobody knew in time.
When I ask what changed, teams usually point at a competitor. It’s broader than that. Your shopper picks up their standards somewhere else. They’re used to checkouts that take seconds, deliveries that beat the estimate, returns that take one click, and a site that seems to know what they want before they’ve finished typing. AI has made all of that cheap to build, so it’s everywhere, and they don’t know or care which of those came from a competitor, a marketplace or an airline app. They only know it’s possible. So the bar sits outside your category, at the best experience that shopper has had anywhere that week.
They’re also making up their minds before they arrive. They ask an assistant which product suits them, have it weigh you against the alternatives, and start the whole journey there instead of in a search engine or on your site. Bain’s Consumer Lab put the share of US consumers using generative AI for product comparison and recommendations at around 30%, and that reading is already the better part of a year old. Whatever the number is this morning, the comparison happens before a shopper reaches you, and you’re not in the room for it.
Neither of those changes shows up in the places you’d look. When the numbers move and someone goes hunting for the why, they reach for what they already own: session recordings, funnel analytics, A/B results, click tracking, thousands of support conversations. All of it has value. But every one of those records is behavior. It’s security camera footage: you can watch someone pick a product up, turn it over and put it back, and the tape never tells you why. It won’t tell you what they wanted, what they compared you against, or what made them give up. And all of it comes from people who reached your site in the first place. The ones who compared you inside an assistant and never arrived aren’t in your funnel at all.
Read More: SalesTechStar Interview with Ilyas Kurklu, Co-founder and CEO of Replenit
So what you know about your shopper has to stay current: as they change, as competitors ship, as the market moves. You already work that way everywhere else in the business. You watch sales daily. You review pipeline weekly. You optimize campaigns in real time. Nobody sets strategy on six-month-old financials. What you understand about the person using your site sits underneath most of what you decide: what you build next, how you price it, how you position it, how you talk about it. That understanding deserves the same treatment.
AI is what makes a current read on the shopper possible, though not in the way most of the conversation suggests. Much of that conversation is about using AI to skip the asking altogether: predicting behavior, generating synthetic personas, summing a customer up into a neat-looking pie chart. That’s still guessing at the why from what you already have, only faster and with a lot more confidence attached. AI earns its place when it gets you closer to real shoppers, letting you hear from people more often, figure out what has changed, and put that in front of whoever is designing the checkout or writing the brief in seconds rather than weeks. You have to be right, not feel right.
This is the shift we’re building Askable around. Instead of commissioning a study each time a question comes up, interviews run continuously and what shoppers say collects into a live repository the team can ask. Ours holds more than 40,000 hours of interviews with consumers. In retail that means AI-moderated sessions with real shoppers, week in, week out. (Real is doing a lot of work there. Fraudulent participants answer just as confidently and nobody notices until the decision ships.) So before you touch the checkout, you can ask what shoppers expect from one now, and see how that has shifted since the last time you looked. The team shipping next month’s release is working from what shoppers said recently, rather than from a picture somebody drew a year ago.
Retail doesn’t need more dashboards telling teams what happened after the fact. It needs a current read on the shoppers who made it through, and on the ones who left. Get that right and the checkout holds, the redesign lands, and your analytics stop being how you find out you were wrong.
About the Author of this Article
John Goleby is the CEO and co-founder of Askable, a human intelligence company. He built it on a bet that looked unfashionable for years and now looks obvious: the best products, and the best models, start with real people.
About Askable
Askable combines participant recruitment, study building, moderation, analysis and reporting in one platform, so most teams replace a recruitment tool, a testing tool and a research repository when they move.
Read More: You Have Cloned Your Voice. Now Your AI Is Making Cold Calls
