One skill that actually matters right now is analytical thinking.
(Most folks used to call this skill “critical thinking”. However, I’m observing a culture shift away from using the word “critical” in certain contexts. I won’t belabor the point or details here, but I will note (1) my aim remains to clarify not confuse, and (2) for my purposes the word “analytical” is more accurate anyway—not correct, accurate.)
It’s in everyone’s best interest to improve our analytical thinking skills when using AI.
Today, I’ll focus on the particular analytical thinking skill of critiquing human questions and answers as well as associated AI inputs and outputs, whether individually (by one’s self) or collectively (in/with teams).
I asked Claude AI to suggest ways to improve my analytical thinking skills.
Here’s part of what it said:
Be Specific About What You Actually Need
The more precisely you describe your goal, the less I have to guess — and the less back-and-forth is required.
State Your Constraints Upfront
Tell me your format, length, audience, and use case at the start, not after I give you something wrong.
Tell Me What You Already Know / Have Tried
This prevents me from restating things you already understand and lets me go deeper faster.
Give Me the Whole Context at Once
Rather than drip-feeding information across many messages (each of which I process fresh), front-load the relevant details. A single well-structured message is far more efficient than five short ones.
Ask for the Format That Serves You
I default to a “reasonable guess” about format. Override it:
Use Positive AND Negative Instructions
Tell me what you want and what you don’t want.
Scope the Task to What You Actually Need Right Now
Asking me to “write a complete business plan” in one go often produces something broad and shallow. Asking me to “draft just the executive summary for a sustainable packaging startup, 200 words” produces something immediately usable.
Break big tasks into focused sub-tasks.
Iterate Surgically, Not Broadly
When you want a revision, be specific about what to change:
Tell Me Your Role in the Situation
“I’m the manager trying to give feedback” vs. “I’m the employee receiving feedback” completely changes what a useful answer looks like.
A Note on Efficiency Specifically
…the single highest-leverage habit is front-loading constraints. Every clarifying round I have to run costs you time and compute. A 30-second investment in writing a more complete prompt routinely saves 3–5 message exchanges. Think of it like a good work order — the more complete the brief, the less rework.
Reading this better-outputs list, I realize that “analytical thinking” when it comes to AI means
• knowing what you need/want,
• knowing what you don’t need/want,
• knowing why, and
• giving AI those three pieces of information in the form of an input/prompt/question.
Sounds obvious.
Easy to forget, though—particularly in moments when I need AI to “just work.”
So I’ll make today’s post even more practical; here are examples of inputs, prompts, and questions that I feed AI.
Specificity
“I’m a certified personal trainer in the United States with qualifications to teach Pilates. What Pilates flow of exercises would address ~80% of a male student’s myofascial pain? Give me a rough estimate of time needed for warm up, exercise, and cooldown.”
State your constraints up front
“Write this draft at an eighth-grade reading level.” “Keep it under 10,000 characters.” “I need this as a bulleted list that I can paste into a slide.” “I’m privacy specialist, so skip the basics.”
Tell AI what you already know / have tried
“I’ve already read Alex Hormozi’s ‘$100M Offers’. What does it miss?” “This web app prototype works, but it’s too slow. I already tried removing a feature. What else?”
Format
“Give me only the explanation—no disclaimers.” “Just the final answer, not your reasoning.” “Walk me through your logic step by step.” “Give me three options, ranked by cost.”
Positive + Negative Instructions
“Draft my business plan executive summary. Make it persuasive but not sales-y. No exclamation points. No phrases like ‘game-changer.’”
Scope to Task
“Draft a 200-word response to the following question for a grant application: ‘What problem are you solving? How do you help your customers with your product or service?’”
Iteration
“The second paragraph is too formal. Make it conversational. Keep everything else.”
Role designation
“I’m the entrepreneur with the Wealth Dynamics Creator profile searching for my ideal Supporter profile candidate.”
Since Claude AI emphasized efficiently specifically, then I will too.
Sometimes, I don’t know what I know, know what I don’t know, know what I need, etc. In such situations, I add a statement at the end of any input/prompt/question I give AI:
Ask clarifying as well as inversion thinking questions as needed.
This one statement has rendered for me WAY more useful outputs.
Try it and let me know how it works for you.