The Floor Staff Didn’t Get Replaced By AI, They got Undermined By It
Image Credit: Getty Images
Picture what each side actually holds. The system on the handheld knows the catalogue cold: every dimension, every refresh rate, every stock figure, every price in real time. The associate knows the person standing there, that they keep glancing at the soundbar, that they’ve mentioned a small flat twice, that they’re nervous about spending this much and want someone to tell them it’s alright. Both kinds of knowledge are real. Only one of them got wired into the technology.
This is the design failure, and the precise shape of it matters. Most in-store AI was built as an answer engine pointed at the product. Ask it about the product, and it answers about the product. It was never pointed at the customer, and it was never pointed at the associate’s judgment, because those were the hard parts and the spec lookup was the easy one to ship. So you end up with a brilliant catalogue, a human standing next to it, and nothing in between.
A Printout of What the Shopper Already Knew
Watch the gap in motion. A customer asks a real question, the kind that brought them in, and the associate reaches for the device, reads back what the screen says, and hands over information the customer could have pulled up on their own phone in the car park. Most of them already did. Around nine in ten shoppers now use their smartphones to compare prices and research products while standing inside the store. The associate has just become a slower version of a search the customer already ran.
And here’s the moment that should bother anyone running stores. The shopper watches the associate look it up. In that small gesture, the trust moves, and it doesn’t move to the associate; it moves to the screen. Once the screen is the authority, the human is friction, a step to get past on the way to a decision. You’ve taken the one thing the physical format can offer, a person worth listening to, and turned them into a worse interface to a database. The whole reason for the trip quietly evaporates.
Consumer electronics has already run this experiment in public. A decade ago, Best Buy was being hollowed out by showrooming: shoppers came in, spent real time with a blue-shirted associate, then bought the same product cheaper on Amazon before they reached the car park. Hubert Joly, who took the chain over in 2012, watched it happen on the floor and described customers leaving empty-handed because they were sure the price online was lower. The store had turned into a place to look, not a place to buy.
Best Buy clawed it back with two moves, and neither was a better screen. It matched online prices, so the phone lost its one advantage, then put real money into associates, more of them, trained deeper in new categories, until talking to a person was worth the trip again. Joly’s name for the shift was turning showrooming into showcasing. By early 2017, the company was posting a third straight year of comparable-store sales growth. The screen was never the asset. The person was.
Image Credit: Getty Images
The tool was built to answer the wrong question
None of this came from a decision to hollow out the floor. It came from scoping. When a retailer buys an in-store AI tool, the requirements that make the cut are usually the ones that are easy to specify and easy to measure. Surface the spec. Check the stock. Quote the price. Suggest the attachment. Each one is legible on a dashboard, so each one gets built.
Yet the parts that actually decide whether a store visit was worth making don’t fit that template. How do you write a ticket to read the customer’s hesitation, or know when the cheaper model is the honest recommendation, or carry what this shopper told us last visit into this one? You can’t, not easily, so it doesn’t get scoped, so the tool ships without it. The result is a system optimised for the measurable half of the job and blind to the half that was the point. We’ve done this before, in every field that automated the easy thing first and called it progress.
What the screen can’t know, and the associate can
The fix starts by being honest about which knowledge belongs where. The screen is unbeatable at the things that are written down. The associate is unbeatable at the things that are not, the read of the room, the trust, the judgment call. A well-designed system doesn’t ask those two to compete. It feeds one into the other.
That’s what clienteling, done properly, is supposed to be. Not the associate reciting specs back at a customer, but the system quietly surfacing what the human can’t hold in their head: that this shopper bought a TV from you eighteen months ago, that the thing they’re touching pairs with it, that they’re a repeat customer worth a small courtesy. In-store, most shoppers are effectively anonymous the moment they walk in, even though their journey crossed your website on the way to the door. The technology’s real job is to un-anonymise them for the one person who can use that, and then get out of the way.
Getting it right
So the design principle is simple to state and apparently hard to buy. AI on the floor should make the associate someone the shopper trusts more than the screen, not less. Everything follows from that one test.
Brief the associate before the shift, not in front of the customer, so they walk the floor already knowing the new range rather than reading it aloud from a handheld. Point the system at the customer, not just the catalogue, so what surfaces is context the human can act on rather than a spec the shopper already saw. Leave the judgment calls with the person, the honest downsell, the sensible exception, the courtesy discount, and let the software advise rather than decide. Then measure the thing that matters, which isn’t lookups per hour. It’s whether the shopper ends up trusting the human in the aisle. Done well, that relationship can influence a sizeable share of monthly store sales, which is the part the finance side should care about. Even the vendors selling this technology will tell you, when pressed, that AI is meant to enhance the human connection rather than replace it. The trick is building it as if you meant that.
What the floor is for
The retailers who come out of this ahead won’t be the ones who put the most screens on the floor. They’ll be the ones who remembered what the floor was for. A store is an expensive way to sell a television unless the person inside it is worth the visit, and right now, a lot of stores are paying for the building and undercutting the only thing in it that the website can’t copy.
The technology isn’t the enemy here. A handheld that knows live stock is a gift, and no associate wants to go back to guessing in the stockroom. The question is only ever what you point it at. Aim it at the catalogue, and you get a faster search the customer didn’t need. Aim it at the customer, through the associate, and you get the one retail experience that still justifies the drive. The tools will keep arriving, each one helpful-looking on the day it ships. Whether they build the person up on the floor or quietly undermine them comes down to who’s paying attention when they do.
SOURCE: TWICE