🔥How to Prompt: The Definitive Guide for Investors, Operators & Builders

Access our most powerful resources here👇

👋 Hi, I’m Andre and welcome to my newsletter Data Driven VC which is all about becoming a better investor with Data & AI. ICYMI, check out some of our most read episodes:

• Top 10 prompts for startup sourcing

• How to scrape alternative data sources

• 16 VC Pitchdecks that convinced LPs to invest

• 312 Family Offices that actively invest in funds & startups

• How to Build an MCP Server for Investors: A Step-by-Step Guide

---

Welcome to another edition of our Sunday “Resources” stream where we share our most valuable data & resources across four rotating formats:

1. 30 Hottest Startups of the Month (September’s list here)

2. Top Downloaded Resources from The Lab #1 (this is today!)

3. State of the Market (see August episode here)

4. Top Downloaded Resources from The Lab #1 (see “how to turn Google Sheets into your competitor radar” here)

For 1. and 3., we collaborate with best-in-class partners to ensure you get the highest quality data. For 2. and 4., we leverage our ever-growing product portfolio and share selective snapshots of the most sought-after resources from The Lab.

Resources Start Compounding

In past “Resources” episodes we shared our top prompts for startup sourcing, screening & due diligence, and deal winning & closing, and various lists with active 312 family offices, 59 pension funds, 1513 angel investors, and 997 accelerators you should know.

Access these and more resources like our 50+ masterclasses, automation templates, Notion templates, copilots, and more via The Lab.

Why You Cannot Afford to be an Average Prompter

I continue to get very positive feedback on our prompt library but what surprises me most is that so many readers still struggle to maximize value from ChatGPT & co by prompting yourself. Therefore, I decided to share all my learnings & best practices of prompting below!

In this episode, we’ll cover:

1. The Basics of Great Prompting

2. Key Prompt Writing Frameworks

3. Advanced Prompting Techniques

4. Special Hacks to Improve the Output

5. Summary + More Resources

This guide is designed as a practical manual and covers the full spectrum of what makes a great prompt. By the end, you’ll have everything to stop being an average prompter and get the most of ChatGPT, Gemini, Claude and others.

Let’s jump in!

---

Part 1: The *Basics* of Great Prompting

Specificity Wins

• Bad: “Tell me about SaaS.”

• Good: “Compare European vertical SaaS companies in healthcare vs. logistics, founded after 2018, that raised Series A in 2024. Output: table with founders, funding, core differentiation, and top customers.”

👉 Rule of thumb: If you wouldn’t ask it in a partner meeting, don’t ask it in a prompt.

Structure the Output

Always define:

• Format: table, bullets, executive summary.

• Depth: overview vs. deep dive.

• Audience: investor, customer, founder, LP.

• Length: word count, paragraph count, slide count.

👉 Example: “2-paragraph summary in LP memo style.”

Iteration Is the Real Superpower

Think of prompting as a loop, not a one-shot:

1. Draft initial request.

2. Inspect output.

3. Refine constraints.

4. Push deeper.

👉 This mimics how you develop an investment thesis.

Anchor With Examples

AI learns better when you show it samples.

• Provide a sample memo for style.

• Provide a mock-up table for format.

• Provide a tone reference (“write like Stratechery, not like a PR press release”).

👉 Helps the model to limit the option room.

Layer Perspectives

Ask the AI to role-play as:

• A founder defending their strategy.

• A competitor critiquing weaknesses.

• An LP testing your thesis.

• A customer evaluating ROI.

👉 This surfaces blind spots early.

Progression: Information → Insight → Action

Most people stop at information retrieval.

• Level 1: “Summarize fintech trends in Europe.”

• Level 2: “Rank trends by probability of collapse in 3 years.”

• Level 3: “Translate into 3 investment theses with example startups.”

👉 The edge lies in pushing one step further than others.

---

Part 2: Key Prompt Writing *Frameworks*

The RICCE Framework (Role – Instruction – Context – Constraints – Examples)

This is one of the most reliable scaffolds for precise, high-quality prompts.

• Role: Who should the AI “be”? (analyst, founder, LP, lawyer)

• Instruction: What exactly should it do? (summarize, analyze, brainstorm, critique)

• Context: What background does it need? (company details, market conditions, user persona)

• Constraints: What boundaries should apply? (word count, tone, format, exclusions)

• Examples: Show what “good” looks like (sample memo, outline, or tone guide).

Example Prompt: “You are a VC analyst (Role). Create an investment memo (Instruction) on a seed-stage AI infra startup raising Series A (Context). Limit to 500 words, bullet point format, no jargon (Constraints). Use the style of the Sequoia scout memo attached below (Example).”

👉 Best for structured outputs like memos, reports, or research summaries.

COTAR Framework (Context – Objective – Task – Action – Result)

Designed for problem-solving and analysis where reasoning matters.

• Context: Set the stage with relevant background.

• Objective: Define the end goal clearly.

• Task: Specify the concrete work the model should do.

• Action: Instruct how to approach the task (step by step, with comparisons, etc.).

• Result: Define the expected output format or deliverable.

Example Prompt: “Context: We are analyzing the Series B SaaS market in Europe. Objective: identify top 5 investment opportunities. Task: compare companies by traction, team, and market size. Action: think step by step, then rank them with justification. Result: provide a ranked table with scores for each factor.”

👉 Best for evaluations, rankings, or decisions where the model needs to “think out loud.”

P-R-A-T Framework (Persona – Request – Audience – Tone)

A lighter framework ideal for communication prompts (emails, LP updates, blog posts, tweets).

• Persona: Who’s speaking? (VC partner, founder, thought-leader, marketer)

• Request: What’s the specific writing task? (draft, rewrite, summarize)

• Audience: Who is it for? (LPs, founders, Gen Z users, enterprise buyers)

• Tone: How should it sound? (formal, persuasive, optimistic, urgent, witty)

Example Prompt: “Persona: VC partner. Request: Draft an email update about our new investment in Company X. Audience: LPs. Tone: crisp, confident, professional.”

👉 Best for tone-sensitive writing — where the how matters as much as the what.

Summary: When to Use Which Framework

Together, these 3 frameworks give you scaffolds for every major use case:

• RICCE → structured research/analysis

• COTAR → problem-solving & reasoning

• PRAT → communication & writing

---

---

Part 3: Advanced Prompting *Techniques*

Delimiters & Structured Formatting

Why it matters: Keeps instructions and content separate, ensuring clarity.

How to apply it:

• “Summarize the following text in 3 bullet points. Text: ‘<insert article here>’”

• “Return results in JSON format with keys for ‘hook’, ‘insight’, and ‘CTA’.”

Step-by-Step Reasoning

Why it matters: Forces the model to “think out loud” instead of jumping to a shallow conclusion.
How to apply it:

• “Think step by step: first identify the issue, then evaluate options, then recommend the best based on feasibility.”

Decomposition (Breaking Big Tasks Down)

Why it matters: Large, complex asks overwhelm AI. Breaking them into smaller chunks leads to depth and accuracy.

How to apply it:

• Step 1: “List 20 climate tech startups.”

• Step 2: “Cluster into 4 categories.”

• Step 3: “Analyze risks for each.”

Role Assignment (Expert Simulation)

Why it matters: Assigning roles activates domain-specific heuristics.

How to apply it:

• “Act as a Gartner analyst forecasting AI adoption.”

• “Act as a CFO evaluating this deal.”

Perspective Forcing

Why it matters: Neutral tone = generic output. Specifying a viewpoint drives sharp insights.

How to apply it:

• “Take the perspective of a skeptical LP.”

• “Write as a founder pitching early believers.”

Constraint-Based Refinement

Why it matters: Narrowing boundaries sharpens clarity and creativity.

How to apply it:

• “Explain this to a high school student using only metaphors.”

• “Write a 50-word summary in active voice, no jargon.”

Multi-Output Generation

Why it matters: One answer = one angle. Multiple versions expand options.

How to apply it:

• “Generate 3 versions of this thesis: conservative, optimistic, contrarian.”

• “Propose 5 alternative headlines with varying tones.”

Iterative Refinement

Why it matters: The real power is in layering — draft → refine → sharpen.

How to apply it:

• “This is too long. Trim to under 100 words.”

• “Now rewrite with stronger verbs.”

Self-Critique & Adversarial Testing

Why it matters: AI sounds confident, even when wrong. Asking it to critique itself reduces errors.

How to apply it:

• “Provide your answer, then highlight possible weaknesses.”

• “Give me 3 counterarguments to your conclusion.”

Style & Tone Transfer

Why it matters: The what matters, but so does the how. Tone determines memorability.

How to apply it:

• “Rewrite this founder update in the style of a Sequoia memo.”

• “Recast this blog post in the voice of Stratechery.”Part 5: Common Mistakes to Avoid

---

Part 4: Special *Hacks* to Improve the Output