ALDI: an assistant that does not become another job

Every new system promises to give employees time back. Then come the login, training, another window, three notifications and a feedback form. Staff become freer mainly in the presentation. A lean store operation is therefore a useful honesty test for any AI proposal: does it remove work, or merely rename it?

ALDI's store telephone information explains that employees focus on serving shoppers, while store and offer information is available online. That is a published explanation, not my judgement about staffing levels. The following is an independent assistance scenario, not a claim about an ALDI deployment.

Recurring questions also reveal information gaps that should be fixed outside the assistant. If several customers ask the same question during a shift, perhaps the store needs a better sign or label, not a more powerful model. This is one of my favourite possible automation outcomes: discovering that the next useful investment is printing two understandable sentences.

The sensible starting point is a recurring question with a verifiable answer: opening hours, returns, category locations and offer conditions. AI retrieves a short answer from approved sources, showing when they were updated. For an employee, this should be one screen rather than fifteen conversational exchanges while a customer watches technological progress unfold.

Urgency matters, and a conversational interface does not understand it merely by sounding helpful. A spill needs immediate attention; a question about a future assortment may wait. The system should suggest a sensible action queue without notifying everyone about everything. Otherwise it adds new noise to old noise, after which management wonders why people have stopped reading notifications.

Time saved should appear in the actual shift

The shift itself should reveal the answer through time to resolve a routine question, repeat interruptions and incorrect-answer frequency. Check essential operational tasks as well. If the assistant helps one customer while delaying replenishment or checkout work, we have simply moved the burden into a neighbouring process and given the movement a more modern name.

A pilot could cover a limited question set during comparable shifts. A supervisor checks initial answers; employees flag mistakes with one action. The store also needs a working fallback. Connections disappear and devices run out of power. Routine service must not become an exercise in waiting for a restart.

Recording personal conversations or scoring employee moods is unnecessary for this task. A useful work tool should not quietly become a surveillance system. My final question is straightforward: does somebody find it genuinely easier to finish their shift? If yes, the technology has earned its place. If not, it is another sophisticated request to do more.