Knitwell Group's COO on Why AI in Stores Starts With a Hello

According to AlixPartners’ Consumer Sentiment Index, a survey of 9,000 US consumers, service as a shopping priority increased 34% year over year. Price fell 13% over the same period. That shift happened during a year defined by inflation anxiety, tariff uncertainty, and widespread cost pressure on the American consumer. Patrick Walsh, COO of Knitwell Group, the portfolio behind Ann Taylor, Chico’s, Talbots, LOFT, White House Black Market, and Soma, was not surprised. “AI won’t have a cigarette with you,” he said, referencing a sign he had spotted in a shop window the night before. The consumer coming into a store in 2026 is more informed than ever, and what she wants from that visit is not a digital assistant in human form. She wants someone who knows her name and understands why she is there.

Knitwell operates thousands of stores, tens of thousands of associates, and millions of customer interactions across its six brands. Walsh’s session at The Lead Summit 2026, moderated by Sonia Lapinsky, Partner, Managing Director and Head of Fashion Retail at AlixPartners, was one of the more operationally grounded conversations of the event. It covered what AI-assisted retail looks like when it is being built inside a real organization, what has worked, what has not, and where the incentive structures quietly undermine the strategy.

The Store Has Already Changed

Before discussing AI, Walsh made a point that reframes the entire conversation about what stores are for. At Ann Taylor today, only 35 out of every 100 in-store transactions involve a staff member selling an item to a customer. The other 65 are buy online pick up in store, eCommerce returns, order from store, and ship from store. The role of the physical location has already shifted. So has the labor model required to run it well.

Walsh described this as a distinction between operating energy and selling energy. Both matter, but they require different people and different organizational thinking. The associate best suited to managing a high volume of omnichannel fulfilment tasks is not necessarily the same person who builds a meaningful client relationship in a fitting room appointment. Knitwell has been restructuring its store labor model around that distinction, and developed two new performance metrics to support it: cost to serve and productivity rate. Both capture the full omnichannel workload of a store rather than measuring output purely against traditional sales-per-labor-hour benchmarks.

The Talbots AI Pilot

The most specific case Walsh shared was an AI pilot running inside Talbots, which he described as the Knitwell brand with the strongest existing client selling culture. Over the past decade, Talbots associates have written more than five million individual handwritten thank-you notes to customers. That relationship depth is the brand’s commercial foundation. It is also, Walsh argued, exactly the right environment in which to test whether AI can genuinely extend that culture rather than flatten it.

Knitwell built a proprietary client-selling tool called the Concierge. For Talbots specifically, the team added an AI component now in active pilot. The mechanics are straightforward. Before a major brand event, such as a friends and family sale or a seasonal launch, the tool generates a contact list matched to individual associates. Previously, out of a potential list of 750,000 individual customer contacts, associates could realistically reach 250,000 to 300,000 by phone, email, or text. With the AI component, the team reached 92% of the full list. Traffic increased sharply.

Crucially, the outreach did not feel automated to the customer. The AI drafts personalized messages based on each customer’s history, preferences, and prior interactions with their associate. The messages go out in the associate’s voice. A Talbots customer who has been shopping with the same associate for years receives a note that reflects that relationship, not a generic promotional email. “We know her family and her backstory,” Walsh said. “What we don’t want is the future to stop reflecting the whole DNA of the brand.”

The rollout will not happen uniformly across all six brands. Walsh was deliberate about that. Talbots is a natural fit because of its relationship culture. Chico’s has a similar client-selling orientation and is next in line. Soma, where the fitting appointment is the gateway to a lifetime customer relationship, requires a different approach. Walsh acknowledged the brand is not yet sure how AI-assisted outreach maps to the intimacy of a fit experience. “We’re going to take it at the pace that each brand feels comfortable with.”

When Incentives Break the Strategy

Walsh’s sharpest observation came from a personal experience. His wife and daughter were shopping recently when a store associate interrupted their purchase to offer 25% off if they booked an appointment. The discount eroded the sale margin. The appointment booking had nothing to do with the customer’s actual experience. The corporate team, however, would see the appointment metric and conclude the initiative was working.

Walsh used this to make a broader point about AI implementation in retail. The technology is only as useful as the incentive structure surrounding it. When corporate teams design metrics and field teams optimize for those metrics rather than for genuine customer outcomes, the result is behavior that looks like progress from a dashboard and feels wrong on the shop floor. “The incentives are an important part of the pilot strategy,” he said. Getting them right is not an afterthought. It is the precondition for the pilot producing signal rather than noise.

The Human Connection Is Still the Conversion Driver

Lapinsky closed the session by returning to the Consumer Sentiment Index finding that opened it. Service is rising as a priority precisely at the moment when AI is making brands smarter about their customers. Walsh’s position was that those two facts are not in tension. They are complementary. AI makes associates more informed, more efficient, and more able to reach more customers meaningfully. But the conversion still happens in a human moment.

“If we don’t have that hello, can you tell me your name, then we’re nowhere,” Lapinsky said. “The stores are nowhere.”

For Knitwell, the work of the next several years is building the infrastructure that gets associates to that hello faster, better informed, and more confident. The AI is the enabler. The relationship is still the product.