Journal / Kirill Naumov
A practical look at frontline AI in retail: what Sam’s Club announced, where manager training matters, and how to test better decisions rather than merely faster reports.
By Kirill Naumov ·
The familiar retail promise goes like this: give store managers a clever tool and the operation will become cleverer. Sam’s Club offers a more interesting question. In September 2025 it said it was putting enterprise-grade ChatGPT tools in the hands of frontline leaders. Its examples were concrete: a club manager could examine a P&L more quickly, another could spot local products, and another could identify a seasonal pattern before adjusting staffing and promotions. Those are company-reported examples, not independently measured business results. But they reveal where the real work lies: in the judgment that follows the answer.
This is an independent reading of Sam’s Club and Walmart public announcements, not an account of a client project. If you run a retail chain, the useful question is not “Should we give every manager a chatbot?” It is “Which decision will this help someone make, and how will we know it was a better decision?”
Saving time on a P&L is useful. Yet a manager still has to decide whether lower sales mean a stock problem, a local competitor, a seasonal change or an error in the data. If the assistant offers a confident explanation without showing its basis, it can make a weak decision happen faster. A good frontline AI pilot therefore begins with one recurring management decision: perhaps where to put colleagues at the busiest hour, which category needs a floor check, or which service issue deserves escalation.
For that decision, write down the evidence a manager normally uses, the data that may be late or incomplete, the options they are allowed to choose and the point at which they must ask another person. Then test AI against real, anonymised past cases. Did it identify the same issue an experienced manager found? Did it flag an uncertainty that the manager initially missed? Did it invent a neat story from a messy spreadsheet? Speed belongs in the scorecard, but only after decision quality.
Walmart separately announced that its US frontline and office-based associates would gain access to tailored AI certification and training in 2026. That is a training commitment in the company’s own words, not proof that every employee has completed it. A generic “how to prompt” class would miss the difficult part of frontline work. Managers need practice questioning a recommendation, checking an underlying number, protecting employee and customer information, and explaining a decision to a colleague who was not in the room.
One useful exercise is deliberately awkward. Give a manager a suggested staffing change, a promotion forecast and a late stock update. Ask what they would verify before acting. Then change one fact: a key item is out of stock, or a regional holiday shifts demand. Do they still follow the recommendation? Good training rewards a justified refusal or a request for more information, not just a polished prompt. The goal is confident ownership of a decision, not dependence on a persuasive screen.
Walmart’s June 2025 announcement described AI task guidance, translation and an upgraded associate assistant. It said early users estimated that planning time for one overnight-stocking workflow fell from 90 to 30 minutes; this is an estimate reported by the company for an initial use case, not evidence that every store saw the same result. Such tools also need a correction path. Can a colleague tell the system that the plan ignores a delivery delay, a blocked aisle or a member need? Does the next recommendation use that feedback, and can a manager see what changed?
For a smaller chain, a credible first pilot might involve five to ten stores, one repeatable decision and a short weekly review with managers and colleagues. Compare it with the existing process. Track decision errors, rework, time spent, staff confidence and service effects. Keep a record of cases where AI should have said “I do not know.” If the only number reported upward is time saved, the organisation may simply have learned to make mistakes at higher speed.
The useful lesson from Sam’s Club is not that AI replaces experience. It is that experience becomes more valuable when people can interrogate information quickly, act with clear authority and notice when the tool is wrong. Retail AI works when the manager remains a manager—and training prepares them for that responsibility.
Explore the English AI for Retail workshop → · Discuss frontline AI training
Primary sources: Sam’s Club on AI for frontline leaders, Walmart on associate training, and Walmart on associate tools. Examples and estimates are company-reported, not independently measured outcomes.