A user submits the message: 'I require assistance with my order from last Tuesday, #98234, as it hasn't reached me.' To prepare this for a Large Language Model (LLM), the application needs to infuse pertinent order information into the prompt. What is the OPTIMAL method for augmenting the context for such an input?
Answer & explanation
Correct answer: D
WHY D is correct: Effective prompt augmentation starts by identifying key fields and intents from the user input (order ID, issue type, date reference), then retrieving dynamic user-specific data from authoritative systems and injecting that real-time context into the prompt. This grounds the LLM's response in accurate, current information rather than general parametric knowledge.
WHY NOT A: Passing the raw message without augmentation forces the LLM to answer from training knowledge alone. It cannot access live order status data, so responses will be generic and potentially fabricated.
WHY NOT C: Retrieving similar past order issues introduces unrelated prior cases. The user's specific order #98234 may differ entirely from the retrieved examples, leading to an irrelevant or misleading response.
WHY NOT B: Static example order data does not reflect the real status of the user's specific order. Assuming uniform status across all orders would produce factually incorrect answers for customers.