Lowe’s and the retail AI companion: train the next question, not just the first answer

A customer has a leaking tap, a photograph on a phone and a perfectly reasonable request for a quick answer. The associate may know plumbing, may work in garden, or may simply be the nearest helpful person. This is where an AI companion can be genuinely useful: not as a tiny oracle in a handheld device, but as a fast way to find the right next question.

Lowe’s says it launched Mylow Companion to associates across more than 1,700 stores in May 2025. According to the company, the tool provides product details, project advice and inventory information on sales-floor devices, and is intended to support customer service and onboarding. It also says the tool accepts conversational prompts and voice-to-text. Those are company-reported deployment details; they should not be read as a universal promise that every answer is correct or suitable for every job.

The interesting lesson is not that an assistant can answer a question about fertilizer or a fitting. It is that a colleague who has just received an answer now has a new job: establish whether this is the right customer, product, store, stock position and level of risk. A fluent answer is the start of judgement, not a substitute for it.

Build the learning moment around the second question

Most prompt practice stops too early. A learner asks, ‘Which seal do I need?’ and receives an apparently useful response. The better exercise is what follows: Is the item actually in this store? Does the customer’s installation match the assumptions? Is there a safety or warranty condition that calls for a specialist? What is the official product or policy source? That sequence turns a chatbot demo into retail judgement.

A strong simulation gives the learner a plausible first answer and then changes one fact. The requested part is out of stock. The customer has given an incomplete measurement. The online description conflicts with the packaging. The learner must explain what they would check and how they would phrase the pause. ‘Let me make sure this fits your situation’ is not a failure of service. It is often the service.

The training content must make its sources visible. Product knowledge, local inventory and safety guidance do not have the same owner or refresh cycle. If an assistant combines them, the interface should show what came from where and when it was last updated. The colleague should be able to open the original source, not merely admire a polished summary.

Let business problems choose the pilot

Lowe’s describes a 2025 AI Day where business leaders put forward challenges, with use cases from functions including supply chain, merchandising and store operations. That is a sensible principle for any learning programme: start with a recurring business problem, not with a generic lesson called ‘how to use AI’. The best first use case is boring in the most productive way—frequent, time-sensitive, measurable and safe enough to test.

For a store team, one pilot might be project-advice conversations in a single department. Prepare a small approved knowledge base, a clear escalation map and scenarios that contain incomplete information on purpose. Managers review a sample of conversations for source use, appropriate referrals and corrections—not just speed. If the assistant makes the colleague sound confident while sending customers home with the wrong thing, the pilot has learned the wrong lesson very efficiently.

Measures should include time to a verified answer, correct escalation, avoided rework, customer follow-up and colleague confidence after a real shift. Add a small qualitative review: which questions repeatedly fail because the data is missing, contradictory or not local enough? That list is more valuable than a dashboard full of prompt counts. It tells the retailer what to fix in products, process or content.

An AI companion earns trust when it helps people become more careful without becoming slower. The useful retail superpower is not pretending to know every aisle. It is knowing how to get from a hurried question to a checked answer—and when the honest next move is to bring in a human expert.