Santalucía Seguros streamlines insurance agent support with GenAI-powered virtual assistant
Santalucía Seguros, a historic Spanish insurance company, implemented a Generative AI virtual assistant to empower its agents with fast, accurate access to information on insurance products, procedures, and coverages. Available via Microsoft Teams, the assistant uses natural language and is accessible 24/7 across devices, allowing agents to respond instantly to customer queries, thus expediting issue resolution and increasing sales productivity. The system is architected on a Retrieval Augmented Generation (RAG) framework, deployed within Santalucía’s Advanced Analytics Platform running on Databricks and Microsoft Azure. Key innovations include ongoing ingestion and embedding of up-to-date documentation into vector stores, and robust LLMOps for integrating, validating, and rolling out updates without sacrificing response quality or reliability. Databricks Mosaic AI Model Serving enables secure access and management of both Azure OpenAI and 3rd party LLMs, governed by robust workflows and integrated CI/CD. This ensures agility and scalability while maintaining compliance and data privacy. The scalable solution drives greater agent productivity, accelerates customer response times, and supports business growth with measurable improvements in user experience.
- Organization
- Santalucía Seguros
- Industry
- Insurance
- Location
- Spain
Reported outcomes
Strategic outcomes
Primary read
Use case focus
Showing 3 of 3
- 1Virtual Agent Assistant for Insurance Product Support
- 2GenAI-Powered FAQ and Document Search for Insurance Agents
- 3Continuous Embedding Ingestion Pipeline
- Agents required immediate, reliable access to extensive product documentation.
- Inefficient, slow manual searches delayed customer service responses.
- Need to support agents with up-to-date information in a compliant and secure way.
- Continuous delivery of new AI model releases without risking response quality.
- Built a Microsoft Teams-embedded virtual assistant using Generative AI and RAG.
- Continuous ingestion and embedding of insurance documentation into vector stores for rapid LLM-powered retrieval.
- Deployed Databricks Mosaic AI Model Serving for scalable, secure, and abstracted LLM integration.
- Utilized Azure OpenAI and Databricks MLflow pipelines for robust LLMOps, versioning, and CI/CD.
- Customer questions answered in seconds, significantly increasing agent productivity.
- Accelerated insurance product sales via instant, accurate responses.
- Improved customer satisfaction with faster, always-available service.
- Continuous, high-quality updates to models and content without interruption to service.
Architecture
The solution is based on a RAG framework, with documentation ingestion and embedding into Databricks-powered vector stores. LLMs (including Azure OpenAI) are served via Mosaic AI endpoints, secured and governed with Azure Key Vault; MLflow-powered CI/CD enables the regular deployment of model updates while maintaining robust access controls and compliance.
Implementation partners1
Sources & evidence1
AI-generated summary. Verify important details with the linked sources before relying on this case.