AI Cross-sell: recommendation system, related products, upsell, cross-selling

What this solution does

AI Cross-sell: Recommendation system based on YandexGPT for automated upselling

Missing recommendations on a product page results in lost revenue. A customer finds the desired item but fails to see relevant add-ons, leading to a single-item purchase instead of multiple items. Manually configuring links for thousands of products requires weeks of effort, becomes outdated instantly, and is rarely completed.

AI Cross-sell solves this problem. The YandexGPT neural network independently analyzes the catalog and generates meaningful recommendations for each product. The system operates without rigid rules or manual intervention, enabling 24/7 sales.

Business benefits:

Increased average order value. Customers see relevant add-ons at the moment of selection and add them to their order without switching to competitors.

Free sales channel. Recommendations appear on every product page and in the shopping cart 24/7 without requiring an advertising budget.

Team time savings. There is no need to manually configure links or update them when new items arrive; the neural network handles these tasks in the background.

Always up-to-date offers. When the catalog is updated, simply re-upload the index, and the recommendations are automatically recalculated.

The module automatically identifies relationships between product variations, including sizes, colors, and offers, correctly mapping them to parent items.

The neural network generates three types of recommendation blocks for each product, which can be enabled or disabled independently:

Complementary products. Items logically suited for add-on purchases, such as a case for a smartphone, a cable for a laptop, or a filter for a coffee machine.

Products with similar prices. Alternatives from the same category that serve as a backup in case of rejection: if the current item does not fit, the customer sees similar options.

Products with similar attributes. Items from the same brand, made of similar materials, or intended for the same purpose.

Flexible configuration without coding. The prompt formulation for the neural network can be adjusted directly in the module settings to adapt the logic to the specific catalog. No code modification is required.

Seamless design integration. The module renders recommendations using the standard Bitrix component (bitrix:catalog.section), eliminating the need for template rework and ensuring full style consistency. On Aspro solutions, blocks appear as part of the Aspro system; on other designs, they integrate organically into the existing layout. The same template already used on the site is applied.

The component supports two placement options.

The module generates personalized recommendations powered by artificial intelligence. On product cards, the system suggests items relevant to the product currently viewed by the buyer. In the shopping cart and during checkout, recommendations are selected based on cart contents, prioritizing the most expensive items.

How it works

The product catalog is exported to the Yandex AI Studio search index. A background agent of the module sends data to the YandexGPT neural network, ensuring no additional load on the website. The algorithm analyzes characteristics, prices, and categories to generate three types of recommendations. Results are saved and displayed within the existing site design.

Full control in the admin panel

All processes are transparent and manageable without developer intervention. Features include tracking processing progress, searching for recommendations by product, manual adjustments, recalculating for specific items, and viewing request logs for prompt debugging.

System requirements

The system requires 1C-Bitrix (main module version 18 or higher), information blocks, and a catalog. A Yandex Cloud account with access to AI Studio is required (services are billed separately).

General iT can install and configure this module and check how it works on your website.