A customer puts a ten-year-old food processor on the counter. The bowl gear is broken. The rating plate is illegible, the model is no longer sold, and the sales assistant does not know which part to order. They would have to leaf through the exploded views in several manufacturer catalogues, compare similar references and call a wholesaler. The shop advises the customer to try elsewhere.
The legal framework favours repair
For household appliances, small IT and telecommunications equipment, and displays and monitors, the manufacturer must ensure that spare parts are available while the model is on the market and for at least five years after the last unit is placed on the market. Once the information has been provided, the manufacturer supplies the parts to the sellers and repairers who request them within two months (Code de la consommation, article L111-4).
The European right to repair directive, which Member States had to transpose by 31 July 2026, also requires that parts and repair information be made available on reasonable terms (Directive (EU) 2024/1799).
Parts therefore remain available, often for years. What the shop lacks is a way of knowing which one fits the appliance on the counter.
What the system does
The sales assistant photographs the appliance, the broken part and whatever remains of the rating plate. The system then asks a few targeted questions: the colour of the bowl, the number of speeds, the shape of the shaft. Based on the answers, it:
- identifies the brand, range and most likely model, along with the neighbouring models it could not rule out;
- finds the part in the manufacturer’s exploded view;
- indicates whether it is available from the retailer’s suppliers, and with what lead time;
- proposes a compatible substitute part when the original part no longer exists.
The check: compatibility has to be proven
A part ordered by mistake means a return, a second visit from the customer, and often a loss of their trust. A second system, separate from the one that identified the part, checks the proposal before it reaches the sales assistant:
- the selected model links to the manufacturer documentation in which it appears;
- the proposed part appears in the exploded view of that model, with its reference;
- a substitute part is proposed only if manufacturer or supplier documentation establishes its compatibility;
- when the identification remains uncertain between two models, the proposal says so and indicates the question that would settle it.
A proposal without documentary confirmation is presented as uncertain. The sales assistant knows whether they can order or whether they must first check on the appliance.
What remains the shop’s responsibility
The system identifies. The decision to take on the repair rests with the shop, which knows whether it is within its means and whether the price makes sense given the value of the appliance. Diagnosing the fault itself, when it is not visible, remains the technician’s job.
Where to start
One product family is enough to get started, the one with the most refused requests. You will need:
- the catalogues and exploded views of the main brands in that family;
- the prices and lead times of the retailer’s parts suppliers;
- the history of repairs already handled, where it exists;
- a sales assistant or technician who knows these products, to assess the proposals.
The system works from these documents, without connecting to the till or the ERP. The result is measured by the revenue generated from parts and labour on requests that the shop used to send elsewhere.
Other uses of AI in retail, and a detailed scenario, are described on the page AI applied to retail.