LLMs work best in dialogue, not as “one-and-done” tools. Think of them as smart interns—or pocket-sized consultants. Give them roles, constraints, and step-by-step tasks. Then push back, clarify, and iterate.

1. AI as Researcher & Analyst

Prompt

You are a market researcher and industry analyst. 
Analyze the latest market trends and buyer preferences in [your category]. 
Identify key segments, pain points, and purchasing behaviors. 
Deliver insights in a table, with notes on assumptions and confidence levels. 
Take your time and double check your work

Why it works

• Assigns the model a role.

• Builds research into positioning and targeting.

• Surfaces pain points and buying triggers for sharper messaging.

Pro Tip

Follow up with: “Pressure-test the top three insights. What would falsify each?”

2. AI as Strategist & Planner

This uses Chain-of-Thought prompting, which asks the model to reason step by step.

Prompt

You’re a senior research strategist. 
Help me plan discovery for [product/offer]. 
Let’s think this through step by step: 
1) Define 3–5 research objectives. 
2) Recommend methods (qual/quant) per objective. 
3) Draft 12 core research questions. 
Before you propose, ask me clarifying questions

Why it works

• Forces structured reasoning.

• Encourages the model to clarify assumptions.

• Produces more rigorous and testable plans.

Pro Tip

Ask: “Trim this to a 2-week sprint. What’s the 80/20 version?”

3. AI as Product Strategist

Adapted from work by Wharton professors Ethan and Lilach Mollick.

Setup
Give the model product details first, or ask it to summarize reliable web sources. Confirm accuracy before continuing.

Prompt

Using the confirmed product info: 
1) List potential customers, why they’d buy, and pick one ICP. Bold assumptions. 
2) Draft a 3-email demand-gen sequence for that ICP. Include subject lines and cadence. 
3) Propose a single-page launch strategy (hero, proof, offer, CTA). 
4) Generate production-ready copy and a component list for a landing page. 
5) Outline a 2-week social campaign with 6 posts across LinkedIn, X, and Instagram. 
Pause after each step for feedback before continuing

Why it works

• Shows the complexity LLMs can handle.

• Builds in checkpoints so you can adjust midstream.

• Produces integrated outputs across email, web, and social.

Guardrails That Improve Output

• Give roles and scope: “Act as [role]. You can do X, not Y.”

• Make assumptions visible: “Bold assumptions and propose how to verify them.”

• Ask for evidence: “Cite sources or show confidence levels.”

• Use constraints: budget, time, or channel—on purpose.

Use the Frameworks

Prompts like these aren’t scripts—they’re frameworks. Use them to turn LLMs from passive tools into active thought partners, and you’ll unlock more insight, strategy, and execution than you thought possible.