Hermès: unavailable should not become a theatre of scarcity

'Very few left' is a remarkable phrase. It can describe stock, or the customer's remaining patience. When every absence becomes a mystery, people struggle to tell whether an item is unavailable, the information is old or they are being encouraged to want it with slightly greater intensity.

Do not promise what the data cannot support

The official Hermès online store provides context for discussing selection and availability in luxury retail. I am not describing internal allocation rules or making claims about the brand's conduct. This is an independent proposal for clearer AI-assisted communication in situations involving uncertain availability.

The first mechanism generates answers only from verified states: available in the chosen channel, unavailable, timing unknown or clarification required. AI translates a technical status into ordinary language. It cannot append an estimated date of its own. 'Soon' is still a promise, however elegantly it is phrased.

The second mechanism is a customer-requested notification. A person chooses which change matters and can cancel that request. Contact follows an actual event, not an automated need to maintain engagement. Repeated reminders that absolutely nothing has changed are unlikely to make waiting feel more considerate.

Third, alternatives should not overwrite the original wish. An adviser may offer another option, but AI should not frame an unrelated purchase as a condition for obtaining the desired item. Ask permission, explain the difference and accept refusal. Wanting precisely what is unavailable remains a perfectly reasonable customer position.

Prepare a reply for the case where no new information exists. Keep it brief and avoid manufacturing anticipation. It may look too ordinary for an AI project, but it tests whether the system can preserve the truth when it has no pleasant update to offer.

Politeness must not conceal uncertainty

The system needs current market and channel data, its update timestamp and an escalation route to a person. If the status is uncertain, the answer should say so. It must not invent waiting lists, priority arrangements or reservation promises absent from the verified operating rules.

A pilot could cover one type of availability enquiry without changing sales policy. Compare existing replies with the new process under equivalent stock conditions. An operator sees the source status alongside the suggested wording and approves dispatch. Any invented promise triggers immediate review rather than waiting for a cheerful end-of-pilot presentation.

My three measures would be accuracy against availability records, repeated enquiries caused by an unclear answer and complaints about unfulfilled promises. Sales alone cannot judge this intervention. Explaining an absence honestly may preserve a relationship even when no transaction happens that afternoon.

AI marketing is particularly useful when it helps a business communicate an inconvenient truth clearly. No theatrical urgency, no predictions on behalf of the warehouse, no invitation to decode hints. Respect can sound very simple: this is what we know, this remains uncertain, and this is the next useful step.