To face this scenario, Made implemented the Íris Digital Agent, a solution based on Generative Artificial Intelligence, using Amazon Bedrock and large language models, integrated with a vectorized knowledge base with official Eucatex technical content.
The solution was designed with an architecture composed of three specialized multimodal agents, who act in an orchestrated manner to deliver accurate and contextualized responses to the user:
– Stock and Availability Agent
Responsible for consulting updated information on product availability, lines, variations and stock status, allowing for more assertive responses aligned with Eucatex's commercial reality.
– Specialist Problem Diagnosis Agent
Focused on identifying pathologies, application failures and problems technical, this agent interprets texts, images and audio sent by users, analyzing surfaces, contexts of use and reported symptoms to support technical diagnoses in a safe and consistent way.
– Product Recommendation Agent
Specialized in the entire Eucatex catalogue, this agent crosses the technical diagnosis, the context of use and the characteristics of the products to recommend the most appropriate solution, providing the user with clear, objective and technically based final guidance.
The agents operate with defined technical criteria and business rules, ensuring adherence to the sector's technical standards, product leaflets and Eucatex's internal guidelines. The solution also features advanced visual understanding, integration with regulatory bases and an administrative platform that allows you to monitor interactions, monitor indicators such as NPS, identify recurring doubts and continually evolve the knowledge base.
Designed with a scalable, secure and governed architecture, Íris is prepared for integration with multiple digital channels, such as the web and WhatsApp, supporting the future evolution of Eucatex's digital strategy.