The Research Happened Before the Customer Walked In
Most retail transactions still happen in a physical store, and most of those still depend on an associate at some point in the path to purchase. What has changed is how much work the customer has already done before that conversation starts. Retail Dive's reporting describes shoppers spending a couple of hours or more researching a purchase using AI tools before ever visiting a location, arriving with specific questions, comparisons already made, and a short list narrowed down. The associate is no longer introducing the customer to options, they are finishing a decision the customer mostly made somewhere else.
That shift mirrors what happened when retailer websites first let customers research online before visiting a store, except compressed into a fraction of the time and with far more confident customers. The problem is that most store technology and training programs were not built for this moment. An associate now has, by one estimate in the reporting, less than a minute to confirm or meaningfully add to a decision a customer has already spent hours forming, and most current store systems are not designed to help an associate do that quickly.
Why Handing Associates a Chatbot Is Not the Fix
Nikki Baird, vice president of strategy and product at Aptos Retail, argues that the instinct to solve this by giving associates the same consumer-grade AI tools shoppers use at home is a mistake. Associates are, in her words, way more time crunched than a shopper browsing from a couch, and a tool that produces an unreliable answer in a busy aisle does more damage than no tool at all. Baird says she has seen plenty of in-store technology get stuffed in a drawer specifically because of that unreliability, and once staff lose trust in a tool, getting them to pick it back up is harder than getting it right the first time.
That is a sharp warning for any retail CTO currently rolling out a generic AI assistant to frontline staff under pressure to show fast progress. The tool that works well for a customer doing open-ended research at home is not automatically the right tool for an associate who needs a precise, fast, and reliably correct answer in front of a waiting customer. Those are different design problems, and treating them as the same one is how good AI investments end up sitting unused.
Training Has to Go Past What the AI Already Told the Customer
Garry Heon, senior vice president and sector head of retail and consumer goods at QualityAI, frames the training gap differently: AI will only return answers for what you ask it, which means a customer's AI research is only as good as the questions they thought to ask. An associate's real value is in noticing the question the customer did not think to ask, the complication specific to their situation that no general-purpose chatbot would have surfaced. Heon describes this as the AI bringing the associate to the next level, rather than associates competing with or replacing what the AI already provided.
That reframes the training problem. Associates do not need a crash course in how AI works in the abstract. They need training specific to the complex or expensive products in their store, built around the gap between a generic AI answer and the situational judgment a customer is actually paying for when they choose to shop in person rather than order online. Retailers investing heavily in branded AI tools for customers should be allocating a comparable share of that budget to this kind of associate training, not treating it as a smaller afterthought.
The Upside: Coverage When Stores Are Short-Staffed
The same technology that creates this training gap can also close a different one. Heon points to AI's potential to help associates work competently outside their usual department when a store is short-staffed, giving a clothing specialist enough grounding to field a basic question in electronics or home goods during a busy shift. That use case is lower risk than customer-facing chat, because the output is a prompt for a human conversation rather than a direct answer a customer sees, and it directly addresses a staffing problem most retailers already have.
For a CTO weighing where to spend the next AI budget cycle in-store, this is a more defensible starting point than a general customer chatbot for associates. It solves a concrete operational problem, it carries lower risk if the AI output is imperfect, and it gives frontline staff a reason to trust the tool before you ask them to rely on it for something higher stakes. Build credibility there first, then expand scope once associates have seen the tool work.
Measuring the Right Thing Before the Holidays
Most retailers currently measure in-store AI success by adoption numbers, logins, sessions, queries answered, the metrics that are easy to pull from a dashboard. Baird's and Heon's observations suggest a better measure is how often an associate's answer goes beyond what the customer could already get from an AI assistant at home, since that is the specific value the store visit is supposed to add. A retailer that tracks only usage will keep funding tools associates quietly stop opening, while a retailer that tracks whether associates are adding judgment the AI could not provide will catch a failing rollout before the holidays expose it at scale.
With the holiday season approaching and AI-assisted research already shaping how shoppers arrive at stores, this is the window to run that measurement, not January. A short, focused pilot in one or two departments, scored on whether associates are resolving the question the customer's AI research left unanswered, will tell a retail technology team more about where to invest for 2027 than a chain-wide rollout judged only on how many associates logged in during Black Friday week.
