Guide Thinking, Not Just Asking
Chain of thought prompting: Getting AI to work through problems step by step
Hey Alchemists,
Yesterday I showed you why context isn't optional. How adding real stakes, relationship reality, and constraints turns generic AI responses into strategic ones.
But here's what I didn't tell you: Even with perfect structure and context, AI still gives you first-draft thinking.
Today we're diving into Prompt Crafting Principle #4: Guide Thinking, Not Just Asking.
This is where you stop getting AI's initial reaction and start getting its best work. Where you turn AI from a fast typist into an actual thinking partner.
The Problem with Fast Answers
Here's what most people don't realize about AI: It's optimized for speed, not depth.
Ask AI "How should I price my consulting?" and it'll immediately spit out a response based on the first pricing framework it finds in its training data.
But that's not thinking. That's pattern matching.
Real thinking looks different. It's messy. It considers multiple angles. It works through alternatives before settling on an answer.
That's what we're missing when we just ask AI for answers.
What I've Noticed After Using AI Daily
Here's what I've observed after months of using AI for everything from writing to strategy to technical problem-solving:
AI gives you its first thought, not its best thought.
Ask AI "How do I secure my AWS environment?" and it'll immediately spit out the most common security checklist from its training data. Enable MFA, use IAM roles, encrypt everything all the standard advice you've heard before.
But that's not thinking. That's pattern matching.
I started noticing this pattern everywhere. Ask about cloud architecture? Generic best practices. Ask about security frameworks? The same compliance checklists everyone shares. Ask about technical problems? Surface-level solutions that sound smart but don't actually fit your specific environment.
AI was giving me answers, but it wasn't thinking through my specific technical challenges.
The Breakthrough Moment
The shift happened when I accidentally discovered something.
I was brainstorming content patterns for my SecureAI Weekly newsletter. Instead of asking "What content should I create about AI security?", I was frustrated with getting the same generic topic suggestions and typed out something like:
"I'm trying to figure out content patterns for SecureAI Weekly that actually help people understand AI security risks. Walk me through how to think about this step by step. First, what are the biggest gaps between what people think AI security means versus what it actually involves? Then help me analyze what content formats would best address these knowledge gaps..."
The response was completely different. Instead of generic "write about data privacy and model security" advice, AI actually worked through the psychology of how people misunderstand AI risks, analyzed the difference between technical and business-focused security concerns, considered what formats work best for different audiences, and gave me specific reasoning for why certain content patterns would resonate more than others.
Same AI. Same question. But I made it think instead of just respond.
What Actually Works: Making AI Think
"You are a pricing consultant helping a marketing strategist figure out their rates. Here's the situation: 8 years experience, specializes in B2B lead generation strategy (not execution), targeting companies with $5-20M revenue. Current approach: $150/hour, but working too many hours and clients aren't seeing the strategic value.
Walk me through your thinking process:
1. First, analyze what's wrong with the current pricing model
2. Then consider 3 different pricing approaches for this situation
3. Evaluate the pros and cons of each approach
4. Recommend the best option with specific reasoning
5. Explain how to position this pricing to clients
Think through each step carefully before moving to the next one."
That's when I realized: AI is capable of deep thinking or sort of deep thinking …
You just have to guide the process.
Instead of asking for conclusions, ask for reasoning.
Instead of requesting answers, request analysis.
Instead of demanding solutions, demand exploration.
This is called "chain of thought prompting," and it's the difference between getting AI's first idea and getting its best idea.
How to Make AI Actually Think
Method 1: The Step-by-Step Breakdown
Don't ask: "How do I improve my productivity?"
Try:
"I need to improve my productivity. Walk through your analysis: 1. First, identify the 3 most common productivity problems people face 2. Then analyze which problem is likely affecting me based on this situation: [your current challenges] 3. For the top problem, brainstorm 5 potential solutions 4. Evaluate each solution for: ease of implementation, potential impact, and time investment required 5. Recommend the best starting point with specific reasoning
Take your time with each step. Don't rush to conclusions."Method 2: The Multiple Perspective Approach
Don't ask: "Should I learn Python for data analysis?"
Try:
"I'm considering learning Python for data analysis.
Help me think through this decision:
1. First, argue FOR learning Python (3 strongest reasons for my situation)
2. Then argue AGAINST learning Python (3 strongest concerns or alternatives)
3. Consider a middle-ground approach (what that might look like)
4. Based on this analysis, what questions should I ask myself to make the right decision?
5. What information do I need to gather before committing time to this?
Work through each perspective thoroughly before moving on."
Method 3: The Problem Decomposition
Don't ask: "How do I get better at public speaking?"
Try:
"I want to improve my public speaking skills.
Let's break this down systematically:
1. First, analyze the components of effective public speaking: content, delivery, audience engagement, managing nerves
2. For each component, identify what good vs. poor performance looks like
3. Based on common patterns, where do most people struggle the most?
4. Given my situation [brief description of your experience level], which area is likely my biggest bottleneck?
5. For that bottleneck area, what are 3 specific practice methods I could try?
Think through each component carefully before diagnosing my main challenge."
Why This Actually Works
Here's the thing about AI: It has access to millions of examples of good thinking. You just have to activate it.
When you ask for an answer, AI gives you the most common response to your question.
When you ask for a thinking process, AI actually uses the reasoning patterns it learned from the best thinkers in its training data.
Same intelligence … Different activation method.
The Thinking Prompt Formula
Here's my go-to structure for getting AI to think instead of just respond:
1. Set the Context (the situation and constraints)
2. Define the Process (the steps you want AI to work through)
3. Specify the Output (how to present the thinking)
4. Emphasize Depth (tell AI not to rush)
Example:
Context: I'm a teacher trying to make my online lessons more engaging for high school students who seem distracted during virtual classes
Process:
Work through this in 4 steps: analyze why students get distracted online, identify engagement techniques that work in virtual settings, evaluate which techniques fit my teaching style and subject, recommend 3 specific strategies I can implement this week
Output: Show your reasoning for each step before moving to the next
Depth: Take time to consider different learning styles and attention spans. Don't give me your first thought—give me your best thought."The Real Test
You'll know you're getting AI to think when:
The responses get longer and more detailed
AI starts saying things like "On one hand... but on the other hand..."
You see actual reasoning, not just conclusions
The solutions are more specific to your situation
AI considers downsides and limitations, not just benefits
Your Thinking Homework
Take any challenge you're currently facing whether it's personal development, learning a new skill, career decisions, or solving a problem at work. Instead of asking AI for a solution, ask it to think through the challenge with you.
Use this template:
I'm dealing with [specific challenge].
Help me think through this systematically:
1. First, analyze what's really causing this challenge (not just symptoms)
2. Consider 3 different approaches to addressing it
3. For each approach, identify potential obstacles and how to work around them
4. Recommend the best starting point with clear reasoning 5. Outline how I'll know if it's working
Work through each step carefully. Show your reasoning.Try it. See how different the output is compared to just asking "How do I solve X?"
Whether you're trying to build better habits, learn a new technology, improve relationships, advance your career, or start something new—this approach works.
Because here's what I've learned: AI can think. You just have to teach it how.
Tomorrow: Can You Transfer What You've Learned? - The remix test of taking prompt crafting principles and applying them to completely new situations.
Drop a comment: What's one problem where you keep getting surface-level advice from AI? Let's design a thinking prompt that gets you deeper insights.