Doctolib revolutionizes healthcare customer care with AI-powered RAG system

Doctolib, a leading European e-health company, implemented an advanced AI-powered customer care solution to enhance its support services. Their journey began with deploying Retrieval Augmented Generation (RAG) to power customer FAQs, using GPT-4o through Azure OpenAI Service and OpenSearch vector databases for dynamic knowledge retrieval. The team built robust data pipelines for continuous FAQ embedding and leveraged machine learning classifiers to increase answer precision. An evaluation tool measured key metrics such as context precision, recall, faithfulness, and answer relevancy to optimize the system. With iterative improvements, including prompt engineering and reranking, Doctolib reduced the volume of deflected cases and improved user satisfaction. Key challenges like system latency were addressed through architectural adjustments and model optimization. The article outlines a path toward more sophisticated agentic AI frameworks capable of handling even more complex queries and actions. Limitations of conventional scripted bots were overcome as LLMs (Large Language Models) enhanced response adaptability. The integration of RAG enabled context-aware responses using up-to-date internal documentation. However, the system exposed bottlenecks in handling complex, non-FAQ scenarios, motivating explorations into multi-agent agentic architectures for future expansion. The solution underscores Doctolib’s ongoing development, aiming to further streamline healthcare customer care while providing a scalable and secure support framework that protects user data privacy.