All useful. But I think we’re missing the bigger skill. Learn how to think.
Because we are moving into a world where the possibilities are becoming almost infinite.
Twenty years ago, you might have had three practical ways to solve a business problem.
Today AI can give you 50 before breakfast.
That sounds wonderful until you realize something:
Having more choices doesn’t automatically make you smarter.
Sometimes it just gives you more ways to screw something up.
The scarce resource is no longer information.
We have more information than any human being could possibly consume.
The scarce resource is judgment.
Knowing what makes sense. Knowing what doesn’t.
Knowing when something looks impressive but is actually bullshit.
Knowing when you have enough information to move forward and when you need to stop and investigate.
I’ve always liked the old carpenter’s expression:
Measure twice. Cut once. AI makes that philosophy even more important.
Except now you can measure twenty times for practically nothing.
So why wouldn’t you?
Before I have AI write a bunch of code, I would rather spend the cheap tokens discussing the problem, looking for holes, considering alternatives and figuring out exactly what we’re trying to accomplish.
Then build it.
The same thing applies to business. Thinking is usually cheaper than doing.
Doing something twice because you didn’t think the first time is expensive.
FIRST, FIGURE OUT WHAT THE PROBLEM ACTUALLY IS
This sounds obvious. It isn’t.
People constantly start with solutions.
“We need AI.”
“We need a new website.”
“We need another server.”
“We need to automate this.”
Maybe.
But why?
What problem are we solving?
What does success look like?
What are the limitations?
What do we already have?
What happens if we do nothing?
Sometimes the correct answer is a $50,000 system.
Sometimes it’s a $500 system.
Sometimes it’s a 20-line Python script.
And sometimes the smartest thing you can do is leave the damn thing alone.
That’s judgment.
AI is actually very useful here because you can argue with it before spending money.
Tell it your idea. Then tell it: Find everything wrong with this idea.
Ask another AI.
Then ask one to defend the idea.
Compare the arguments.
You’re not asking AI to make your decision.
You’re using AI to help you make a better decision.
Big difference.
DON’T CONFUSE A GOOD ANSWER WITH A CORRECT ANSWER
This one worries me. AI can be wrong beautifully.
It can give you a perfectly formatted, professional, intelligent-sounding answer that is completely wrong.
Humans have been doing this for thousands of years, so AI didn’t exactly invent the problem.
We’ve all sat through a PowerPoint presentation where everything looked wonderful and nobody asked the obvious question:
Does any of this actually make sense?
The important question isn’t:
“Does this sound right?”
It is: “How do I know this is right?”
Show me the numbers.
Show me the source.
Show me the assumption.
Test it.
Run it.
Try to break it.
Ask what would have to be true for the conclusion to be wrong.
If an AI gives me an answer and I don’t understand enough about the subject to recognize a ridiculous answer, then I have another problem.
I don’t know enough yet to make the decision.
LEARN TO BREAK BIG PROBLEMS INTO LITTLE PROBLEMS
This is another thing AI is teaching people that good managers have known forever.
Don’t attack a huge project as one giant project.
Break it apart. When I look at a large system, I don’t think:
“Build the system.”
I think:Â What are the pieces?
What talks to what?
Where does the data come from?
What happens first?
What happens if that fails?
How do we test each piece?
Which parts can be done independently?
That’s project management.
It’s also how good AI agents need to be managed.
The funny thing is that managing five AI agents isn’t completely different from managing five employees.
You still need to tell everybody what their job is.
You still need somebody responsible for the overall objective.
And you still need to check the work.
AI just happens to work about a million times faster.
Which means it can also make a mistake about a million times faster.
ALWAYS ASK: HOW CAN THIS FAIL?
This is probably one of the most useful habits anyone in business can develop.
Whenever somebody shows me a system and says: “It works.”
My next question is usually: “Okay. How does it break?”
What happens when the database is down?
What happens when the internet is down?
What happens when somebody enters bad information?
What happens if the AI misunderstands something?
What happens if one AI agent makes a mistake and hands that mistake to three other agents?
And the most dangerous one:Â What happens if something fails and nobody knows it failed?
That last one is huge with AI.
A machine that crashes is actually easier to deal with.
You know it crashed.
A machine that quietly gives you the wrong answer and tells you everything is fine is much more dangerous.
That’s why critical thinking matters.
Don’t just test whether something works.
Test whether you can tell when it doesn’t work.
NOT EVERY MISTAKE IS EQUALLY IMPORTANT
Business is risk management.
If ChatGPT misspells something in a draft email, who cares? Fix it.
If an AI accidentally deletes a production database, now we have a different conversation.
If it recommends the wrong headline, annoying.
If it transfers money to the wrong bank account, catastrophic.
So whenever I automate something, I like to think about four things:
How bad can the mistake be?
How often could it happen?
Can I reverse it?
Can I verify it?
The more dangerous the answer, the more human oversight you need.
This is where people make a big mistake with AI.
The goal shouldn’t be: “Remove the human.”
The goal should be:
“Put the human where the human actually matters.”
Let AI do the boring 95%.
Have somebody competent watching the dangerous 5%.
CONTEXT IS EVERYTHING
You can ask an AI a perfectly reasonable question and get a perfectly reasonable answer that is completely useless to your business.
Why?
It doesn’t know enough. What kind of company?
How many employees?
What hardware?
What budget?
What software?
What regulations?
What are we already doing?
What are our priorities?
What did we try last year?
What failed?
That information changes the answer.
This is why I think one of the biggest areas in AI over the next few years won’t just be bigger models.
It will be better context.
Giving the AI the right information at the right moment.
The same applies to people.
I can give two intelligent consultants exactly the same problem. If one understands my business and one doesn’t, I know which answer I’m going to trust.
AND FINALLY: DOES IT MAKE ECONOMIC SENSE?
I love technology. That doesn’t mean every technical solution is a good business solution.
I can probably automate almost anything now. That doesn’t mean I should.
If something takes an employee ten minutes once a month, I don’t need to spend three weeks designing an AI agent to save those ten minutes.
Congratulations. I just spent $10,000 solving a $50 problem.
AI makes building things much easier.
That means we need to become more disciplined about deciding what deserves to be built.
What’s the cost?
What’s the benefit?
What’s the maintenance?
What’s the risk?
What’s the alternative?
How much human time does it save?
How important is that time?
Can a smaller model do it?
Can a script do it?
Can Excel do it?
Can we simply change the process?
That’s business judgment.
And this is why I don’t think the winners of the AI revolution will necessarily be the people who know the most about AI.
They may be the people who know how to use AI while still thinking for themselves.
The tools will change. ChatGPT will change. Claude will change. The models will get better. Agents will get better. Programming may change completely.
But some skills aren’t going anywhere:
Define the problem.
Question the assumptions.
Break complexity into pieces.
Look for failure.
Verify the important stuff.
Understand the risk.
Consider the economics.
Then make the decision.
In other words:Â Measure twice. Cut once.
Except we now live in a world where measuring is incredibly cheap.
So maybe the new rule should be:
Measure ten times. Ask three AIs. Try to prove yourself wrong. Then cut once.
Because in a world of infinite possibilities, the great skill isn’t coming up with another possible answer.
AI can already do that.
The great skill is knowing which answer is worth acting on.
