A financial institution wants to build an internal knowledge base chatbot for its employees. The chatbot must be able to answer complex, domain-specific questions by retrieving information from a large repository of internal documents, and then synthesize a concise, accurate answer based on the retrieved content, rather than just pointing to documents. The solution needs to minimize hallucinations and provide traceability to source documents. Which Azure AI solution pattern is best suited for this requirement?
- AAzure AI Language for Key Phrase Extraction and Text Summarization
- BAzure AI Speech for Speech-to-text and Neural Text-to-Speech
- CRetrieval Augmented Generation (RAG) with Azure AI Search and Azure OpenAI Service
- DDirect use of Azure OpenAI Service with a large language model (LLM)
Show answer & explanationAnswer & explanation
Correct answer: C. Retrieval Augmented Generation (RAG) with Azure AI Search and Azure OpenAI Service
Retrieval Augmented Generation (RAG) is the ideal pattern for this scenario. It combines Azure AI Search to retrieve relevant internal documents with Azure OpenAI Service to generate a concise, accurate answer based *only* on the retrieved content, minimizing hallucinations and providing traceability.
Why the other options are wrong
- A. Key Phrase Extraction and Text Summarization are analytical tools, not generative or retrieval systems for chatbots.
- B. Speech-to-text and Neural Text-to-Speech are for voice interfaces, not for complex knowledge retrieval and generation from documents.
- D. Direct LLM use can hallucinate and lacks traceability to specific internal documents for answers.
Retrieval Augmented Generation (RAG)
An AI pattern that combines information retrieval with generative AI. It first retrieves relevant information from a knowledge base or data source and then uses a large language model (LLM) to generate an answer grounded in that retrieved information, improving accuracy and reducing hallucinations.
- Combines retrieval (e.g., search) and generation (LLM).
- Reduces LLM hallucinations by grounding answers in factual data.
- Provides traceability to source documents.
- Ideal for domain-specific Q&A from private data.
Memory trick: Search the facts, then Speak the truth.