A company needs a chatbot to answer employee questions about company policies. Policies change frequently and the chatbot must reflect updates almost immediately. Which LLM-based approach satisfies this requirement?

Implementing a Retrieval-Augmented Generation (RAG) pipeline using Amazon Bedrock Knowledge Bases is the most suitable approach because it allows the chatbot to access and incorporate the latest policy documents without retraining the underlying LLM. When policies change, the knowledge base is updated, and the RAG system retrieves the most current information to generate responses. Fine-tuning an LLM (Option A) would require retraining the model every time policies change, which is time-consuming and expensive, making it impractical for frequent updates. Choosing a foundation model from Amazon Bedrock (Option B) alone doesn't address the need for immediate policy updates; it still requires a mechanism to integrate external, frequently changing data. Amazon Q Business (Option D) is a good solution for business intelligence and data analysis, but a custom Q App might not offer the same level of real-time, granular control over document retrieval and generation for policy-specific questions as a dedicated RAG pipeline.