Question

A financial services company is building a customer support triage assistant. Incoming support emails must be classified into one of five predefined categories (e.g., 'billing', 'fraud', 'account_access'), and the output must be returned as structured JSON so a downstream workflow can route the ticket automatically. The team has no labeled training data but has a clear written description of each category. Which prompt strategy best fits this use case?

Accepted Answer

Use a zero-shot prompt that includes the category definitions and explicitly instructs the model to respond only with JSON matching a specified schema.

Answer

Fine-tune a base LLM on the five categories before deployment, since classification always requires supervised training.

Answer

Use an open-ended free-text prompt asking the model to 'describe what this email is about' and parse the prose response with regular expressions.

Answer

Use a chain-of-thought prompt that outputs its full step-by-step reasoning as the final response for each email.