With support from AWS, Grendene tests AI to support Melissa's sales force

The project for the development of the new virtual assistant is being prepared by the Grendene, AWS, and MadeinWeb teams. Melissa franchise manager, Rafael Tremarim, explains that one of the goals is to bring the e-commerce experience to physical stores, offering customers product recommendations related to what they are buying. To do this, the company will use the Charla AI, developed by MadeinWeb.

Grendene's open innovation analyst, Bianca Nascimento, explains that Charla is being fed with the descriptions of Melissa's last four collections. “She will also receive information about the sales behavior of two of the Melissa clubs, which will allow us to test the assistant in a controlled environment so that it can later be scaled,” she says. These tests are being carried out at Melissa's two owned stores, located in the Ipanema and Leblon neighborhoods, in Rio de Janeiro. “They are our laboratories to test these solutions that will then be driven through the entire network”, reinforces Tremarim.

According to Leonardo Wissmann, commercial head of MadeinWeb, the feeding process should last between four and five weeks. The next phase is data management and creating the Melissa club persona. “The AI must have the characteristics of the brand, ensuring that the assistant will not be an intruder in the process,” he says, explaining that Charla runs in AWS cloud, using a data lake based on Amazon Simple Storage Service (Amazon S3). The service used for model training is Amazon SageMaker and the construct also has Amazon Elastic Compute Cloud (Amazon EC2). The process should become even more effective after migrating the Amazon SageMaker models to Amazon Bedrock, a work that is already in progress.

Gandolfi states that the solution will facilitate the work of Melissa club employees, providing them with accurate information at the right time and converging with the company's sales objective. “Grendene's partnership with AWS and MadeinWeb today is on the company's innovation agenda,” he says, recalling that the initiative will bring digital experience to physical environments.

When the testing phase is complete, the new virtual assistant will be made available to the sales teams of 403 active Melissa clubs in Brazil, the vast majority of which are installed in shopping centers. Tremarim believes that the introduction of new technologies tends to make results appear more quickly. “With this joint movement with AWS, we tend to gain a lot in the assertiveness of what consumers will receive, not only with what they order, but also regarding other products that may be offered together,” he says.

Gandolfi also points out that the initiative is one of the company's first to bring technological solutions, common in B2C relationships, to B2B relations as well. “E-commerce usually provides recommendations for the end customer. We are talking about a value chain, made up of franchisees and sellers who also need to receive these recommendations,” he says.

For the head of retail tech Melissa, Walter Duarte, these recommendations will bring benefits to employees on several fronts. An example is the strategic inventory knowledge and smart redirection, where Charla, using detailed inventory data from the entire network, can offer suggestions for alternative products in the event of a local stock-out, or direct customers to other channels and stores in the chain, using an omnichannel approach.

“We also expect an increase in complementary and cross sales, since the assistant will be able to recommend not only accessories, but complementary products that enhance the customer's shopping experience using product catalog analysis, encouraging additional sales together with the shoes chosen,” he explains.

The new assistant should also provide highly personalized recommendations, based on each customer’s purchase history and individual preferences, improving their experience. “This not only encourages customer loyalty, but it also fosters frequent returns, resulting in a constant increase in sales over time,” he defends.

Finally, Duarte also highlights the possibility of identifying inventory trends and strategies through customer buying pattern and behavior analysis. “It will be possible to identify which products are most likely to be sold, allowing the store to direct its resources to inventory and marketing strategies for products that are in high demand and following current market trends,” he explains.