Categorize and enrich millions of product sheets, then improve research

What was done

In an interview with the Journal du Net published on January 8, 2025, Antoine Pierart, digital director, and Thibault Vigié, director of strategy and transformation, indicates that a dozen boxes of general AI use have been scaled up, on a catalog of 15 million offers and more than 27 million unique visitors per month. The first project was the optimization of the sheets: robots check and correct the catalogization, then the work extended to titles, descriptions, faces and photos. Between a well-categorized sheet and a sheet that is not, "accelerated click rate is very strong", a double-digit growth which exact value is not communicated; "Millions of product sheets" have been enriched annually since 2023. On the search engine, the general AI is used in particular to categorize requests and to integrate the semantic association, improvements then "not visible by the end client".A "gift engine" was tested on a sample of customers during the holidays. The recommendation cars integrates the navigation history and a better understanding of the catalog. The customer service boxes (ten main reasons, answers in the customer space) were "started up" at the time of the interview, so they are not presented here as already generalized. A Data and AI Factory controls the transition to scale; the data are on Google Cloud, the customer experience of the site on Microsoft Azure.

Before you start

Tool cited by the source: General AI (Fnac Darty). Gather the information the prompt asks for. Strip confidential data that is not needed and check your tool’s access terms. Features and plans can change.

The method, in three steps.

1. 01

   Gather the context

   Gather the information described in the prompt.

2. 02

   Adapt the prompt

   Copy the prompt below, replace the bracketed items, then paste it into the AI assistant you use. The cited tool (General AI (Fnac Darty)) is not necessarily available to the public..

3. 03

   Reread before using

   Reread the result, correct it if needed and keep the final decision. Caveats are detailed below.

Your starting point.

Replace the bracketed items, then adapt the proposal to your situation. This prompt is written by the iadujour desk It is not a quotation from the source.

Product sheet to be checked or enriched. Big data (current title, category, beg characteristics): Consignee:

1. Proposed a categorization and title, limiting yourself to the features provided.
2. Writes a clear description, without superlative unwrapped.
3. Lists missing fields (compatibility, dimensions, legal information) to be completed by a human.
Do not invent any technical characteristics.

What deserves your attention