Nike: first ask where the person actually wants to run

When somebody says they want to run, that might mean a first lap around the block, regular training or finally using sportswear somewhere other than an airport. Opening the most technical catalogue immediately tells us a great deal about the catalogue and very little about the person standing in front of it.

Nike provides a Running Shoe Finder. The existence of that tool is the documented starting point. The conversation design below is my independent proposal for AI in retail sales, not an account of how Nike's tool works. I want better questions, not a faster route to the most expensive box.

Establish the purpose before explaining the technology

I would start with where the customer expects to run, how often, and what previously felt comfortable or uncomfortable. Then establish budget and practical constraints. A beginner should not need to pass a vocabulary test about foam before receiving useful advice. Product expertise belongs on the retailer's side of the conversation.

AI turns those answers into a short task profile and suggests two relevant models to the associate, explaining the trade-offs. Not ten cards sharing the word 'comfort', but differences in intended use, price and what to assess during a fitting. The customer tries them, and their experience outranks a beautifully phrased recommendation.

If the purpose changes to include substantial everyday walking, the shortlist changes too. That must not automatically mean a higher budget. The system should be capable of explaining that a simpler option already meets the requirements. Otherwise we have built a checkout with literary ambitions, rather than a useful assistant.

Keep retail advice within retail

The inputs are verified model descriptions, current prices and availability. Questions about pain, injuries or treatment belong with an appropriate professional, not the sales model. It must not diagnose gait from a short video or promise injury prevention. Collecting health histories to improve a basket recommendation is neither necessary nor sensible.

A pilot could cover everyday running footwear with a small team. Product specialists review the initial recommendations; associates then record why customers accepted or rejected a shortlist. That feedback improves the conversation. It is not a demand that every colleague become a perfect reader of human intentions.

The fitting gives us a place to test the proposal: examine consultations with a clearly identified purpose, expectation-related returns and conversion after fitting. Comparisons need similar stock and seasonal conditions. Average order value is interesting, but cannot be the only judge: an expensive unsuitable pair can cost the retailer the next visit.

AI in sports retail becomes useful when it helps someone articulate what they actually want to do. We are not selling a professional athlete's biography in a cardboard box. We are helping someone choose equipment for their next real session. That modest, recognisable usefulness is a good place to build trust.