Quick Overview
The problem: AI can remove useful mental effort along with useless busywork.
The risk: If AI routinely makes the first interpretation or decision, you get fewer chances to practise those skills yourself.
The better model: Think first, use AI second, then think again.
Offload freely: Formatting, transcription, organisation, repetitive analysis, and admin.
Keep ownership: Problem framing, interpretation, trade-offs, opinions, and consequential decisions.
Use AI as resistance: Ask it to find flaws, assumptions, alternatives, and counterarguments.
The caveat: Research is still developing. What matters most is how much thinking your AI workflow still requires from you.The biggest limitation: AI doesn’t understand consequences. Humans do.
The Framework
After experimenting with AI every day for the past few years, I’ve settled on a simple framework.
Think → Challenge → Decide
Stage | Your question | AI’s role |
|---|---|---|
Think | What do I think is happening? | Stay out |
Challenge | What am I missing? | Critique my thinking |
Decide | What will I do? | Present options, but don’t decide |
Use AI Without Losing Your Ability to Think
The cardinal rule about using AI is this: AI is a force multiplier. Here’s what that means in practice:
1. Think Before You Prompt
For work involving interpretation, strategy, creativity, or judgment, give yourself a first pass before asking AI for one. Write three bullets:
what you think,
what you are unsure about, and
what decision you need to make.
Once you have these three written down, you can prompt your LLM to help you strategize:
Prompt:
I have formed an initial view below. Identify weak assumptions, missing information, and gaps in my reasoning. Then give me the three highest-leverage questions I should answer before deciding— ones where the answer could flip my conclusion, not just add detail.
2. Offload Work, Not Judgment
Split the task before you delegate it.
Good to offload | Worth keeping human |
|---|---|
Formatting | Defining the problem |
Transcription | Deciding what matters |
Sorting | Interpreting ambiguity |
Draft variants | Final judgment |
Repetitive research | Choosing between trade-offs |
The boundaries change with expertise which means that experienced people can delegate more because they have a stronger internal standard for judging the output; beginners need more direct practice.
For example: Let AI cluster 100 customer comments by theme. You decide which themes matter commercially and what action follows.
Delegate the repeatable part before the ambiguous part. If you cannot explain why the output is good, you probably delegated too much.
3. Make AI Challenge You
Stop using AI only as a producer. Use it as intellectual resistance. Here’s a prompt that does this:
Prompt:
Act as a skeptical reviewer of [what you're sharing]. Steelman the strongest disagreement an intelligent, informed critic would raise — not a strawman.
Separate your response into:
Factual weaknesses — claims that are wrong or unsupported
Reasoning weaknesses — logic that doesn't hold even if the facts are right
Alternative interpretations — other reasonable ways to read the same evidence
Be specific and cite the exact part you're critiquing. Do not rewrite my work or suggest fixes.
Make sure to ask for counterarguments before improvements. And ask AI to separate facts, assumptions, and inferences.
4. Keep Some Manual Reps
Decide which skills you still want to be good at without AI, then practise them without assistance. For instance, if writing is part of your craft, write some first drafts yourself. If analysis is your job, work through the material yourself first before asking AI what it sees.
A preliminary 2025 MIT Media Lab study found lower neural engagement and weaker recall among participants using an LLM for an essay-writing task than among participants who wrote without tools.
Keep One Skill Completely AI-Free
If I could recommend only one practical exercise, it would be this:
Choose one professional skill and protect it.
If you’re a writer, write something every week without AI.
If you’re a marketer, analyze a campaign yourself before asking AI for an explanation.
If you’re a designer, sketch ideas before generating variations.
When you’re learning, friction is valuable. When you’re producing, efficiency matters. People often treat these two situations as if they’re interchangeable. They aren’t.
Task | Learning mode | Production mode |
|---|---|---|
Writing | Write first, then ask AI to edit | Ask AI to generate a draft |
Research | Read the source material | Ask AI to summarize findings |
Brainstorming | Generate your own ideas first | Use AI to expand possibilities |
Analysis | Work through the problem yourself | Use AI to accelerate repetitive work |
Reading | Read the book | Use AI for review and recall |
When Should AI Stay Out of the Process
AI can produce the draft, analysis, summary, or options. You should still be able to defend the final output.
Before you publish, send, or act, ask:
Do I agree with this?
Can I explain why?
What evidence supports it?
Would I put my name on it?
That final pass is where AI assistance becomes your work rather than something you merely approved.
The goal is to make sure AI expands what you can do without replacing the mental processes you still need.
A useful default is Think → AI → Think Again.
Form a position. Use AI to extend, challenge, organise, or accelerate it. Then take responsibility for what survives.
If you want more practical guides for using AI at work without turning your job into a collection of prompts, subscribe to Practicaly AI. We publish workflows, frameworks, and tutorials for people who want to use AI well without spending their week learning AI tools.
How useful was this guide for you?
💌 We’d Love Your Feedback
If you need any guidance while implementing this, or if something isn't quite clear, feel free to ping the team. We're here to support you and clear things up.
Until next time,
Team PracticalyAI
