A claims pipeline must extract policy number, incident date, and claimed amount from varied document text into a predefined schema. Which task design is most suitable?

1. Open-ended document summarization with no required fields.
2. Semantic similarity search returning only document identifiers.
3. A conversational interface that stores every answer as untyped prose.
4. Structured information extraction with field definitions and output validation.