Melissa launches virtual assistant with generative AI for retail

Melissa, one of Grendene's best-known brands, faced challenges in offering personalized service at scale. Customers expected fast, efficient interactions and accurate product recommendations. However, manual service and support in physical stores had limitations in response speed and personalization of the experience, especially during peak demand.

Grendene also wanted to improve operational efficiency at points of sale, making it easier to consult product availability, personalization and detailed information for customers.

MadeinWeb, in partnership with AWS, developed a Generative AI Virtual Assistant, capable of transforming Melissa's shopping experience at points of sale. This solution leverages advanced Artificial Intelligence technologies and robust AWS infrastructure to provide personalized, fast, and efficient interactions for consumers.

Key components of the solution include:

• AI Virtual Assistant: Developed by MadeinWeb, this assistant interacts with customers via chat, answering product questions, suggesting options based on user preferences, and providing timely stock availability information real.

• Integration with Inventory Systems: The assistant is integrated with Grendene's inventory management system, ensuring that customers and sellers can access product availability accurately and instantly.

• AI-Based Personalization: Using previous sales data and the customer's profile, the assistant suggests personalized products, providing a differentiated shopping experience and increasing opportunities for cross-sell and up-sell.

MadeinWeb implemented the solution using AWS services to ensure the scalability, security and efficiency of the application:

• Amazon Bedrock: We used Bedrock to access the Claude3-Sonnet 3.5 text generation model and the Titan Embedding V1 embedding model. The choice for Claude3 Sonnet 3.5 is due to the complexity of the task to be performed by an LLM, which is to identify and extract information from a knowledge base of Melissa products based on user input.

• Amazon Sagemaker: We used Sagemaker to train the recommendation model. To do this, we use a processing job and deploy it as a container in ECS. The reason we don't deploy it as an endpoint in Sagemker is related to the cost of the solution, which is one of the requirements of the implemented solution.

We worked closely with Grendene to create a tailored solution that combined business needs with consumer expectations:

1. Solution design and architecture: MadeinWeb developed the solution architecture based on AWS infrastructure to ensure scalability and performance, even in periods of high demand.

2. AI model training and customization: The team at MadeinWeb used Amazon SageMaker to train AI models that could understand consumer preferences and intelligently suggest products.

3. Real-time inventory integration: The system was integrated directly into Grendene's inventory management system, providing real-time responses about product availability.

The solution implemented by MadeinWeb brought significant results for Grendene and the Melissa brand:

• Increase in sales: The virtual assistant generated a significant increase in sales by offering personalized recommendations and cross-selling and up-selling effectively.

• Improved customer experience: The agile and personalized interaction with the virtual assistant improved customer satisfaction, providing faster service and efficient.

• Operational efficiency: Integration with the real-time inventory system allowed sellers and customers to have instant access to critical information, optimizing service in physical stores.

• Scalability and performance: The AWS infrastructure allowed the solution to be scaled to different points of sale without losing performance, ensuring a quality experience even during peaks in demand.

MadeinWeb, using the power of AWS technologies, developed an innovative solution that transformed the shopping experience for Melissa, a Grendene brand, at points of sale. The application of generative AI and integration with inventory systems allowed Grendene to offer personalized and efficient service, boosting sales and strengthening customer satisfaction.