Question

A legal services firm wants a generative AI assistant that answers employee questions about the firm's constantly-changing internal policies. Requirements: answers must reflect the latest policy documents (updated weekly), must cite the source document, the firm cannot send data to third-party APIs, and the engineering team has limited ML training resources. Which application design best translates these business requirements into a technical approach?

Accepted Answer

Build a RAG application using an open-source model served on Databricks Model Serving, with a vector index over the policy documents refreshed weekly

Answer

Fine-tune a proprietary hosted LLM on the policy documents each week and query it through the vendor's public API endpoint

Answer

Use a large context-window model and paste the entire policy corpus into every prompt so the model always has current information

Answer

Pre-train a foundation model from scratch on the firm's documents to keep all data in-house and avoid retraining costs