Product descriptions:
Lojas Renner, Brazil’s largest fashion and lifestyle retailer — with over 20,000 employees and hundreds of stores across Brazil, Uruguay and Argentina — is on a mission to “delight everyone” and empower employees with AI-driven insights. Yet, Lojas Renner struggled with siloed data and fragmented systems that slowed innovation and made it difficult to scale their AI initiatives. By leveraging the Databricks Data Intelligence Platform, Lojas Renner built a scalable AI agent factory that automated support calls and reduced response times. This platform also analyzed customer feedback and helped employees across the organization make faster, data-driven decisions.
From fragmented silos to data intelligence
Lojas Renner is Brazil’s largest fashion retailer, employing over 20,000 people across hundreds of stores in South America. Before partnering with Databricks, Lojas Renner’s internal support centers were overwhelmed with thousands of recurring queries from employees about purchase order processes, supplier guidelines and HR policies. “The challenge wasn’t just about responding to the requests,” said Jacson Jacobus, Manager of Engineering and BI at Lojas Renner. “It was about how scattered our information was, locked in PDFs, Word docs and different systems, with no way to unify or query it efficiently. That created silos, slowed us down and diverted time from frontline customer service.”
Fragmented tools across data engineering, data science, and MLOps compounded the issue and limited Lojas Renner’s ability to release AI solutions rapidly. Despite these hurdles, they identified several high-impact AI use cases that promised to improve employee and customer experiences, reduce operating costs and democratize access to insights. These included an Internal Support Agent to automate recurring HR and operational questions; Voice of the Customer Analytics, which transforms chat, voice and social feedback into actionable insights that elevate the shopper journey; a Strategic Insights Hub where analysts can use natural language to uncover insights from customer interactions across voice, chat and WhatsApp without needing SQL expertise; and a Labor Relations Agent that converts legal text and complex Collective Bargaining Agreements into instantly accessible guidance.
Achieving this vision would require a unified data and AI platform to integrate diverse data, enforce governance and scale AI.
Building an AI agent factory for scale
To meet these challenges, Lojas Renner leveraged the Databricks Data Intelligence Platform to unify their entire data and AI lifecycle. Central to their strategy was the creation of an AI agent factory, a reusable, flexible framework that enables rapid design, development and deployment of domain-specific AI agents. This framework approached AI not as one-off projects but as an extendable, scalable system, powering multiple AI agents across HR, legal, customer support and the executive team. “Databricks unified everything, from governance and security to deployment. We built an end-to-end pipeline that just works. It was a game changer,” said Lucas Ferreira, Manager of Engineering and Solutions at Lojas Renner.
The Data Intelligence Platform addressed Lojas Renner’s fragmented data landscape by bringing structured and unstructured data into a lakehouse architecture. Delta Lake formed the backbone as a governed, structured storage layer that supports the implementation of Medallion Architecture. Bronze, Silver and Gold layers refined and prepared data to power analytics and AI models, optimizing costs and ensuring continuous knowledge base updates. Unity Catalog enforced fine-grained access controls and seamless data versioning — critical given the sensitivity of HR and legal data. For ingesting unstructured data, Vector Search is tightly integrated with Delta Lake, automatically syncing updated embeddings for retrieval augmented generation (RAG) querying. “The automatic sync of changes to our Vector Search ensures AI agents serve up-to-date and accurate responses, without any manual overhead,” explained Rhayar Mascarello, Senior Data Engineer and AI Platform Specialist at Lojas Renner.
Databricks AI played a key role in transitioning the AI agents from prototype to production. Through MLflow 3.0 and LangGraph, the team was able to monitor every step of the machine learning lifecycle, from data wrangling and training to deployment and evaluation, incorporating LLM-as-a-Judge methodologies that improved output trustworthiness and drastically reduced hallucinations.
“Generic models don’t know our internal processes, policies or the specific clauses of a Collective Bargaining Agreement. The true value of GenAI for us lies in leveraging the reasoning capabilities of LLMs on our own knowledge base, which experts have curated. This is what ensures relevance, accuracy and, above all, our employees’ trust in the answers,” noted Rhayar.
Databricks’ serverless architecture simplified infrastructure management. Model Serving offered a hassle-free, scalable endpoint for AI agent interaction at low latency. Meanwhile, AI Gateway helped govern API usage, set cost controls and maintain security policies. “Abstracting away infrastructure complexities allowed us to focus fully on delivering business impact,” said Douglas Graciano Littig, Data Engineer Specialist at Lojas Renner.
Databricks also fostered a more collaborative and productive culture across its data and AI teams. Engineers, data scientists and business users collaborated fluidly through shared notebooks. With the ability to test multiple LLMs, including GPT-3.5, GPT-4, Claude and others, within the same system, they could pinpoint the best cost-to-performance trade-offs for each use case. “With Databricks, we sped up collaboration and reduced silos,” explained Jacson. “Our teams move from data ingestion to deployment easily, delivering solutions that truly align with business needs.”
Streamlining retail support and operations with AI agents
With its AI agent factory built on the Databricks Data Intelligence Platform, Lojas Renner has transformed how teams work, unlocking measurable value across operations, support, and strategy. At the support center, the Internal Support Agent received more calls while maintaining a high accuracy rate, significantly reducing the need for escalation and human support. Employees now get instant answers to questions such as “How many vacation days do I have?” or “How do I create a purchase order for suppliers?” Average response times improved, while operational overhead decreased significantly, allowing support teams to focus on more complex issues. “By providing employees with instant, accurate information, we reduce administrative friction and empower them to focus on creating excellent customer experiences,” said Douglas.
The Labor Relations Agent has also delivered a breakthrough in productivity and accuracy. Handling thousands of legal documents, including bulky Collective Bargaining Agreements, the Labor Relations Agent supports over 430 users and responds with legal citations. This task previously required extensive time from consultants and specialists. “The Labor Relations agent decodes complex labor contracts, allowing our teams to act faster and with greater confidence,” said Lucas. Simultaneously, the Executive Insights Agent provides leadership with real-time access to board decisions and discussions, accelerating decision cycles and strategic alignment.
The Voice of the Customer Analytics solution connects data from website activity, chat logs and social platforms, distilling unstructured feedback into actionable business insights. These feed into Power BI dashboards, enabling teams across the company to identify and address pain points, thereby elevating the Net Promoter Score — a critical metric for Lojas Renner’s customer experience. “We’re learning directly from what customers say, driving continuous improvements from product delivery to store operations,” explained Jacson.
The scale and flexibility of the Data Intelligence Platform lay the groundwork for innovation beyond the current use cases. Plans are underway for Lojas Renner to utilize AI for e-commerce visual content generation, creating virtual staging and product imagery, and to launch personalized external customer support agents capable of order tracking and smart return processing through chat channels, such as WhatsApp. “With Databricks, we’re ready to evolve from reactive AI to proactive AI, to predict trends and suggest business actions ahead of time,” said Rhayar.
Quantitative successes paint a compelling picture: improvement in agent retention rates compared to legacy systems, reduction in employee session times when accessing data, and seamless handling of thousands of interactions across multiple countries with consistently low latency. Qualitatively, Databricks has strengthened AI trust by embedding transparency and governance, lowering operational costs and cultivating a culture of collaboration and continuous learning across technical and business teams. “Databricks is a great platform. We built so many things so fast here. It’s truly awesome. This speed is key to achieving our mission to delight everyone,” said Jacson.