Your First Prompts for Advisory Work

A prompt is just an instruction you give an AI in plain language. The difference between a vague prompt and a great one is the difference between a generic paragraph you have to rewrite and a draft you can almost use as-is. This lesson teaches a simple, repeatable structure and walks through your first real advisory prompts.

What You'll Learn

• A four-part prompt structure that works for nearly any advisory task
• How to give the AI a "role" and the context it needs
• Copy-and-adapt prompts for common advisor tasks
• How to iterate on a response instead of starting over

The Four-Part Prompt Structure: Role, Context, Task, Format

Almost every effective prompt has four parts. You do not need headers — a couple of sentences covering each is enough.

1. Role — who you want the AI to act as. "You are an experienced financial planner writing for a client with no finance background."
2. Context — the situation and the relevant facts. "My client is 58, recently widowed, has a $1.2M IRA and a $400k taxable account, and is anxious about market volatility." (See the next lesson on what you can and cannot paste.)
3. Task — exactly what you want produced. "Draft talking points I can use to reassure her without over-promising, and to explain why we are keeping a cash bucket."
4. Format — how you want it back. "Five short bullets, plain language, no jargon, under 200 words."

Put those together and you get something the AI can actually run with. Compare:

• Weak: "Write something to calm down a nervous client."
• Strong: "You are an experienced financial planner. My client is 58, recently widowed, $1.2M IRA plus $400k taxable, anxious about a 9% market drop this quarter. Draft five short bullet talking points I can use in a call to reassure her without over-promising — emphasize her cash bucket covering three years of withdrawals, her plan's stress-test results, and the cost of selling at a low. Plain language, no jargon, under 200 words."

The second one gets you a draft you can lightly edit. The first gets you fortune-cookie filler.

Your First Prompts

Open ChatGPT or Claude in another tab and try these. Replace the bracketed parts.

1. Explain a concept in client-friendly language

You are a financial advisor explaining things to a client with no finance background. Explain what a Roth conversion is, when it can make sense, and the main risks, in about 150 words. Use a simple analogy. Avoid jargon; if you must use a term like "tax bracket," define it in the same sentence.

2. Build a review-meeting agenda

You are my paraplanner. Here are my notes for an annual review with a retired couple, both 67, conservative, drawing $5,000/month: [paste notes]. Draft a one-page meeting agenda with timed sections, the three most important discussion points, and a list of documents I should have on hand. Keep it practical.

3. Draft a follow-up email

You are a financial advisor writing a warm, professional follow-up email after an annual review. We discussed rebalancing toward their target allocation, increasing their emergency fund to six months, and revisiting beneficiary designations. Write a short email (under 200 words) summarizing what we agreed, listing the three action items with who owns each, and proposing a check-in in 90 days. Friendly but not overly casual.

4. Turn numbers into commentary

You are writing a short quarterly note for clients. The model balanced portfolio returned about +2.1% this quarter; bonds were roughly flat, large-cap stocks up, small-caps down. Explain what happened in plain English in under 250 words, remind clients we do not chase short-term moves, and end on a steady, reassuring note. Do not predict future returns or use guarantees.

5. Stress-test your own thinking

You are a skeptical senior advisor reviewing my plan. Here is my proposed recommendation for a client: [paste]. List the five strongest objections or risks a thoughtful colleague would raise, and for each, what additional information I should gather. Be direct.

Iterating Instead of Starting Over

The first answer is a draft, not a verdict. You refine it by talking to the AI like a colleague:

• "Good, but make it warmer and cut it to 150 words."
• "Too much jargon — rewrite for someone who has never heard the term 'rebalancing'."
• "Add a sentence acknowledging that this quarter felt rough."
• "Give me three subject-line options for that email."
• "That tax point isn't quite right — assume she's in the 24% bracket and revise."

Each turn the model keeps the conversation in mind, so you build toward the version you want. If a conversation goes sideways, start a fresh chat with a tighter prompt rather than fighting it.

A Few Habits That Make a Big Difference

• Always give a role and an audience. "Explain to a client" produces very different output than "explain to a colleague."
• Be specific about length and format. "Under 200 words, five bullets" beats "keep it short."
• State the constraints. Tell it not to use guarantees, not to predict returns, not to give tax advice it can't support — it will respect explicit "do not" instructions.
• Ask for options. "Give me three versions" is often faster than perfecting one.
• Review every word before it leaves your hands. This is the rule that never changes.

Key Takeaways

• Structure prompts as Role + Context + Task + Format — a sentence or two each is enough.
• Specific beats vague: state the client situation, the exact deliverable, the length, the tone, and any "do not" constraints.
• Treat the first response as a draft and refine it conversationally — adjust tone, length, jargon, and accuracy turn by turn.
• Always assign a role and an audience, and always review and validate the output before it reaches a client.