Today, in addition to the purchased products, the correct recommendations of the by-products that can be sold with these products are a very important source of income for companies. To achieve this, decision-makers frequently apply association analysis methods. At this point, the traditional apriori algorithm comes to the fore. In this study, a recommendation system is created with the apriori algorithm for two different data sets. In addition, customers who interact with products have been embedded with the word embedding mechanism, which is a natural language processing method that has been frequently used in recommendation systems in recent years. Applying that, similar customer products determined by embedding have been added to the product bundles. It has been observed that expanding bundles via customer embedding is useful in both data sets for some cases, and it gives promising results for future studies.