Prompting for Performance Part I: Why Prompts Matter

From Surface-Level Insights to Deep Analysis

The rapid adoption of generative AI in financial services has set the stage for a new era in investment research and due diligence. Analysts and associates across both the buy-side and sell-side are already experimenting with AI-driven tools to deliver faster insights, streamline workflows, and expand the breadth of analyses. Yet one crucial factor is often overlooked when organizations begin exploring these technologies: the importance of prompting.

In AI terms, a “prompt” isn’t just a simple query. It represents a structured request that determines the quality and relevance of the AI’s output. By refining prompts—whether through short queries or multi-step instructions—investment teams can hone the system’s focus, reduce errors, and uncover non-obvious patterns.

In this 3-part deep dive, we’ll explore why prompting is essential to using AI in an investment research context, outline best practices for crafting effective prompts, and discuss how an organization-wide approach to prompt engineering can amplify returns in investment research and due diligence.

AI’s rising role in finance.

Investment research and due diligence often involve analyzing vast amounts of information under tight timelines. AI systems, especially LLMs, are now capable of summarizing large volumes of documents from data room folders to public filings, extracting insights from them, and even performing preliminary analysis on deals. However, these models don’t automatically know what information is most relevant or reliable for a given investment thesis – they rely on our instructions. As a 2025 report from Deloitte highlights, “prompt engineering” has emerged as an important new skillset for finance professionals.” In other words, asking the right questions in the right way is now a key competency.

The power of a well-crafted query.

A poorly worded prompt might return generic or even misleading output, whereas a precise prompt can yield deep, targeted insights. Researchers at AllianceBernstein have noted that a richer dialogue between human experts and AI leads to better outcomes in investment research. That’s where prompt engineering comes in to craft precise queries that direct LLMs toward accurate, relevant responses. For example, instead of asking an AI “Tell me about Company X,” an analyst might prompt: “Summarize Company X’s revenue growth drivers over the past 5 years and identify any one-time events affecting performance.” The latter prompt provides context and specificity, guiding the AI to deliver a more useful answer.

Understanding AI’s limits and strengths.

Finally, it’s important to remember that AI models generate answers based on patterns in data, not guaranteed truth. Users who ask a vague question will most likely receive a vague answer. Even worse, without proper guidance, an AI might “hallucinate” – confidently produce an incorrect fact or figure. Effective prompting mitigates this by setting clear expectations and boundaries for the answers that can be generated. In practice, this means specifying the scope (e.g., “based on the latest annual report”), the format (e.g., “bullet points highlighting risk factors”), or even the reasoning approach (e.g., “consider step-by-step financial analysis”). Understanding the full breadth of prompt engineering techniques can empower analysts to harness the full potential of LLMs and their hidden reasoning capabilities. In other words, the quality of output from an AI tool is directly tied to the clarity and precision of the input prompt.

In the next part of this blog series, we’ll explore best practices and techniques for financial research prompting to maximize the utility of AI tools in this space.

Check out the follow-up posts in the series here: Part II and Part III