Foundations of Prompting: Input Strategies

This chapter examines foundational input strategies for interacting with generative AI, focusing on zero-shot and few-shot prompting techniques within accounting and finance applications. It demonstrates how the framing of instructions directly influences output quality, format consistency, and analytical precision. Zero-shot prompting is presented as an efficient approach for straightforward or well-defined tasks, while few-shot prompting is shown to enhance reliability, domain alignment, and structural control through illustrative examples. The chapter compares both approaches across dimensions such as reasoning depth, efficiency, output stability, and risk of misinterpretation. Practical scenarios in auditing, financial reporting, compliance review, and investment analysis illustrate when each strategy is appropriate. By emphasizing the balance between simplicity and guidance, the chapter establishes that effective prompting is not merely about asking questions but about designing structured inputs aligned with task complexity and risk sensitivity. These input strategies form the operational foundation for more advanced reasoning techniques introduced later in the book.