Trending
Written by: John O’Connell | The Oasis Group
Artificial intelligence tools, including large language models (“LLMs”) like ChatGPT and Microsoft CoPilot, are rapidly becoming part of the modern wealth manager's toolkit. A recent study from Financial Planning magazine stated that 58% of wealth managers said that they use AI occasionally or more frequently in their work.
Whether drafting client emails, conducting investment research, or summarizing tax strategies, AI can be a powerful assistant. However, just like a junior analyst, its value depends entirely on how clearly you ask it to perform its task.
That process is known as prompt engineering. This article outlines what it is, why it matters in wealth management, the main prompt types, and how to use them to get better results.
What Is Prompt Engineering?
Prompt engineering is the practice of designing effective inputs to guide an AI model's outputs. In short, it’s how you frame your question or request.
A vague prompt like "What’s a good investment right now?" will yield broad, generic advice. But a focused prompt like "List three sectors that historically outperform during rising interest rate cycles and explain the key risks for a client in retirement" generates useful, context-aware insights.
Prompt engineering ensures the AI gives answers that are accurate, aligned with your tone, and relevant to client needs.
Prompt Engineering Reduces Risks
Wealth managers operate in a regulated, detail-oriented environment. When using AI to support your work, effective prompts help mitigate risks like:
-
Bias: Overgeneralization or skewed output without proper context
-
Hallucinations: Fabricated facts or citations. See my recent articles on hallucinations to understand how to recognize them and avoid them.
-
Compliance issues: Inappropriate tone or speculative language
Prompt engineering gives you control. It helps to ensure that AI serves your process without disrupting compliance or your professional voice.
The Six Types of Prompt Engineering
Here are six practical forms of prompt engineering, including pros, cons, and wealth management use cases.
1. Zero-Shot Prompting
Zero-shot prompting involves presenting a direct question or instruction without offering any examples or additional context. It is best suited for quick, general summaries or exploratory drafts that don’t require adherence to a specific structure or tone.
Example: "Summarize today’s market trends."
Pros:
-
Fast and easy
-
Good for idea generation or internal drafts
Cons:
-
Output may be vague or imprecise
-
Not ideal for client-facing use
Verdict: Use for internal ideation, not for final deliverables.
2. One-Shot Prompting
One-shot prompting includes a single example alongside your request to help guide the AI’s tone or structure. This method is especially helpful for tasks like personalized client communication, where mirroring a specific format is important.
Example: "Here’s a client note for Hermoine. Now write one for Harry, who just retired."
Pros:
-
Helps align tone
-
Easy to replicate communication style
Cons:
-
One example may not set a consistent pattern
-
Still needs close review
Verdict: Useful for one-off emails or meeting summaries.
3. Few-Shot Prompting
Few-shot prompting provides the AI with multiple examples to help it learn desired patterns. It is particularly effective for creating investment summaries, planning memos, or any task where consistency across outputs is important.
Example: Provide the AI tool with two summaries of ETFs, followed by: "Now summarize this ETF."
Pros:
-
Consistent tone and format
-
Scales well for recurring tasks
Cons:
-
Requires more setup
-
Long prompts may exceed input limits
Verdict: Ideal for teams looking to standardize high-quality output.
4. Targeted Prompting
Targeted prompting incorporates specific instructions about tone, compliance, or formatting. It is ideal for generating materials that are client-facing or must meet regulatory standards, such as client memos, performance reports, or proposals.
Example: "Write this as a fiduciary advisor, no forecasts, plain language, and include a disclosure."
Pros:
-
Reduces compliance risk
-
Aligns with brand voice
Cons:
-
May sound too scripted if overused
Verdict: Crucial for professional, compliant content.
5. Chain-of-Thought Prompting
Chain-of-thought prompting requests that the AI explain its reasoning step-by-step. This approach is well-suited for financial research and planning decisions, scenario modeling, and tasks that benefit from a clear breakdown of logic and assumptions.
Example: "Explain whether a Roth conversion makes sense, step by step."
Pros:
-
Transparent logic
-
Supports analytical thinking
Cons:
-
May produce longer outputs
-
Requires careful review of reasoning
Verdict: Valuable for internal planning or educational content.
6. Role-Based Prompting
Role-based prompting instructs the AI to assume a specific professional role or perspective, such as a fiduciary advisor or CFA. It is especially useful for drafting advanced planning advice, educational content, or high-net-worth client strategies.
Example: "Act as a CFA advising on tax-efficient giving for a $20M client."
Pros:
-
Adds domain relevance
-
Produces expert-like responses
Cons:
-
Still requires human validation
-
Can sound overconfident or inaccurate
Verdict: Great for drafts but always review for accuracy.
Prompt engineering isn’t about learning code. It’s about learning to ask better questions.
AI becomes more than a novelty when it is used intentionally. It becomes a reliable partner that helps wealth managers communicate clearly, make decisions faster, and deliver higher-value advice for clients at scale.
AI can make a wealth manager more efficient. However, wealth managers should always maintain the HITL, or human in the loop, to ensure that the AI results are reviewed by a human before being used in your business.
Start small. Refine your approach. Then scale what works.