Skip to main content

Prompting for Analysis and Brainstorming

You pasted a spreadsheet into AI and asked "what does this mean?" It gave you a three-page essay that said nothing. Or you asked "help me brainstorm" and got ten ideas you could have thought of yourself in two minutes.

Here's the problem: analysis and brainstorming are not simple writing tasks. AI is not drafting an email. It's functioning as a thinking partner. And the way you prompt it determines whether you get genuine insight or just rearranged obviousness.

This topic covers three use cases: interpreting data and comparing options, using AI for ideation and strategy, and prompting AI to teach you anything.

Making Sense Of Data​

AI can read spreadsheets, compare datasets, and spot trends — but only if you tell it what kind of analysis you actually want.

Interpreting Numbers​

Don't ask AI to "look at this data." Give it the data and a specific question.

Here are monthly sales figures for the last 12 months:

Jan: $82K  Feb: $79K  Mar: $91K  Apr: $88K
May: $95K  Jun: $102K Jul: $87K  Aug: $83K
Sep: $90K  Oct: $97K  Nov: $115K Dec: $108K

What patterns do you see? Identify any trends, anomalies, or
questions I should investigate further.

AI will spot the November spike, the mid-year peak, and the August dip. More importantly, it will suggest what to ask next: was November seasonal? Was July a supply issue? This is the real value — not the summary itself, but the questions it surfaces.

Pro Tip

Paste the data first, then ask the question. If you ask the question first, the AI has to hold it in memory while reading the numbers. Data first, question second is a small ordering change that improves output quality.

Comparing Options​

AI excels at structured comparisons — vendor analysis, product comparisons, strategy trade-offs. The key is specifying your decision criteria.

Compare these three project management tools for a team of 15:

Tool A: Asana ($10.99/user/month, strong task management, no built-in time tracking)
Tool B: Monday.com ($12/user/month, visual boards, good reporting, steep learning curve)
Tool C: ClickUp ($7/user/month, feature-rich, can feel bloated, slower support)

Decision criteria:
- Budget under $150/month
- Must support time tracking (native or integration)
- Team is non-technical
- Need basic reporting for leadership

Create a comparison table and recommend one.
Why This Works

Notice what makes this prompt effective: the decision criteria come before the request for a recommendation. You're not asking AI to pick arbitrarily. You're giving it the lens through which to evaluate. The table alone is useful, but the criteria-driven recommendation is what makes it decision-quality output.

The "What Am I Missing?" Prompt​

This is the most underrated analytical prompt. After any analysis, follow up:

Based on the analysis above, what perspectives, data points,
or angles have I not considered?

AI will routinely suggest things you didn't think of: seasonal adjustments, competitor benchmarking, customer segmentation breaks. It's not perfect, but it's cheap insurance against blind spots.

Brainstorming And Strategy​

"Help me brainstorm" is the second-most-ineffective AI prompt, right after "write me a blog post." Here's how to actually get useful ideas.

Structured Ideation​

Instead of asking for "ideas," ask for ideas with constraints and categories.

We're a small marketing team launching a B2B software product next month.
Budget is tight — under $5K for launch marketing. Target audience is
operations managers at mid-size companies.

Generate 10 launch marketing ideas organized into three categories:
1. Zero-cost ideas (no budget)
2. Low-cost ideas (under $500)
3. Moderate-cost ideas ($500-$2K)

For each idea, include: one-sentence description, expected effort
(low/medium/high), and the one thing that could make it fail.

This structure beats "brainstorm marketing ideas" because it forces specificity. Categories prevent the AI from giving you ten variations of the same thing. The "what could make it fail" clause builds in critical thinking that brainstorming usually lacks.

Strategy As Conversation​

Treat AI as a strategy sounding board, not a strategy generator. The most effective approach is conversational:

Step 1 — Frame the problem:
Our customer retention dropped from 92% to 85% over the last
two quarters. We think it's pricing-related but aren't sure.
What are the possible causes and how would we test each one?

Step 2 — Pick one cause and dig deeper:
You mentioned onboarding experience could be weak. What specific
metrics should we look at, and what would indicate that's the
problem vs. it being a red herring?

Step 3 — Pressure-test the plan:
If we improve onboarding and retention doesn't budge, what's our
next most likely cause and what's our backup plan?
The Multi-Prompt Advantage

Notice this is three prompts, not one. Complex problems benefit from staged prompting — each prompt builds on the previous one. This is different from writing prompts where you typically send one. Analysis and strategy work best as a dialogue.

The Devil's Advocate Prompt​

When you have a plan and want to find its weaknesses before presenting it:

Here's my plan: we're switching from monthly to annual billing
with a 20% discount to improve retention.

Act as a skeptical board member. Challenge this plan. What
are the 5 strongest arguments against it? What customer
segments would this hurt most?

The key here is specifying the role ("skeptical board member"). Without it, AI tends to be agreeable and lists weak objections. Give it permission to be genuinely tough and you get useful pushback.

Learning With AI​

AI is the best tutor available — infinitely patient, endlessly adaptable, and free. Most people use it wrong.

Explain Anything​

The "explain like I'm five" prompt is famous for a reason. But you can go further by specifying your current level and your goal.

I understand basic accounting but have never heard of cohort
analysis. Explain it to me like I'm a business professional,
not a data scientist. Use a concrete example with made-up numbers.
Then tell me three questions I should ask my analyst about this.

This works because it anchors the explanation at your level, demonstrates with numbers instead of abstractions, and gives you follow-up questions. You leave with both understanding and a way to go deeper.

Study Guides And Summaries​

I need to learn the basics of [topic] for a meeting next week.
Create a study guide with:
- 5 core concepts I must understand (with 2-sentence explanations)
- Common misconceptions to avoid
- 3 real-world examples
- A 10-minute practice quiz with answers
Start with the basics — assume I know nothing about this topic.

The practice quiz is what separates this from a passive reading exercise. Answering AI-generated questions forces active recall, which is how learning actually sticks.

The Feynman Technique With AI​

The Feynman technique says you truly understand something when you can explain it simply. AI makes this interactive:

Step 1 — Ask AI: "Teach me how [topic] works in simple terms."
Step 2 — Explain it back: "Here's my understanding: [your explanation]"
Step 3 — Ask AI: "What did I get wrong or miss? Be specific."
Step 4 — Repeat until you can explain it clearly.

This loop is more effective than reading articles or watching videos because it forces you to produce the explanation, not just consume it.

The Universal Analysis Template​

Everything in this topic follows one pattern:

  1. CONTEXT: [What situation/data are we working with]
  2. DATA: [Paste relevant data, facts, or background]
  3. TASK: [What kind of analysis or brainstorming you want]
  4. CRITERIA: [Decision criteria, constraints, or evaluation lens]
  5. FORMAT: [How you want the output — table, list, essay, Q&A]
  6. FOLLOW-UP: [What deeper questions to surface]

Example: Strategy Session​

Context: Planning our Q3 marketing budget allocation.
Data: Q2 spent $12K total — LinkedIn $5K, Google Ads $4K,
  Content $2K, Events $1K. Leads: LinkedIn 60, Google Ads 85,
  Content 40, Events 25.
Task: Analyze cost per lead and recommend how to reallocate Q3.
Criteria: Total budget stays $12K. Must maintain at least 2 channels
  for diversification. Want 15% more total leads.
Format: Table with analysis, then 2-3 specific recommendations.
Follow-up: What assumptions am I making that could be wrong?

Same skeleton you learned for writing in the previous topic, applied to analysis instead. The structure is portable.

One Rule

For analysis and brainstorming, the quality of your input data and constraints determines the quality of output. Garbage in, garbage out still applies — but now you're feeding in facts, not just vague requests.

Try This Now

  1. Analyze something real. Open any spreadsheet or data source from your work — sales numbers, website traffic, survey results. Paste it into AI with the "what patterns do you see?" prompt. See what it notices that you didn't.

  2. Run a brainstorming session. Pick a work decision where you're weighing options. Use the structured ideation prompt with three categories and a failure clause. Compare the AI's ideas to yours.

  3. Learn something new with the Feynman loop. Pick a topic you've been meaning to understand — blockchain, cohort analysis, OKRs. Run through the four-step teach-explain-feedback-repeat cycle. Notice how much faster you learn compared to reading articles.

Good Read​