Zara: a fitting-room customer should not become the stockroom manager

A customer in a fitting room has already done half the store's work. They found the garment, carried it over, changed clothes and decided it almost works. All they need is another size. Sometimes the entire relationship ends there. Getting dressed again to begin a stockroom expedition should not be the price of buying trousers.

Zara offers in-store availability checks, and its guidance explains that associates can help order unavailable items online. Those are documented services. The workflow below is my independent proposal for AI in fashion retail, not a claim about how Zara currently operates.

Turn the question into a task

From the fitting room, a customer requests another size or colour against a product identifier. AI converts the everyday wording into a structured task, asks only for missing information and routes it to an associate. 'We are looking for it' is pleasant language. It is not, by itself, an operating process.

The associate accepts the task and confirms physical availability. Until then, the interface says 'checking', not 'on its way'. If the item cannot be found, the system offers a confirmed next step: another store, an online order or an agreed alternative. It must not silently substitute a neighbouring size to rescue a conversion metric.

I would limit active requests per colleague. Otherwise the technology efficiently creates a queue of promises that nobody can fulfil. A manager sees the workload and redistributes tasks. The customer receives an updated expectation instead of an endless animation designed to suggest that something excellent is happening somewhere.

A stock figure is not a found garment

The system needs a product catalogue, recent inventory, task status and human confirmation. Stale data must be labelled. Fitting-room cameras and body recognition are entirely unnecessary; a product code and a request provide enough information. Preferences should expire with the session unless the customer explicitly chooses otherwise.

A pilot could start in one fitting-room zone, comparing similar trading periods with the ordinary process. Size shortages and team workload need to be recorded because they can explain the result better than the interface. Two scenarios are enough initially: a fulfilled size change and an honest, useful refusal.

The customer will notice before the dashboard does. Track time to physical fulfilment, abandoned requests and aggregate fitting-session conversion. Conversion must not depend on covert observation. Complaints about availability promises are an essential guardrail. More purchases accompanied by more disappointment would be a poor trade, however attractive the first weekly chart looks.

The purpose of AI here is wonderfully ordinary: prevent interest from evaporating between the fitting room and stockroom. Better sales do not always require a more persuasive speech about fashion. Sometimes they require bringing the right trousers while the person is still willing to try them on.