A company built an Amazon Bedrock AI assistant to suggest products, but its responses are often generic or irrelevant. Which prompt-engineering technique is most likely to improve the relevance of the assistant's replies?

Few-shot prompting provides the model with examples of desired input-output pairs, along with clear instructions and domain-specific context. This helps the model understand the expected format and content of relevant responses, significantly improving accuracy and reducing generic replies. Chain-of-thought prompting is for complex reasoning, not primarily for relevance, and hidden reasoning steps would not override explicit instructions. Changing conversational style doesn't inherently improve relevance. Zero-shot prompting provides no examples, making it less effective for improving relevance compared to few-shot when the model struggles with generic outputs.