You know why you clicked here – you want to understand the skill in question. Congrats, you already know the answer: knowing what you want.
But is that really true? The title was vague, yet you’re here anyway. You came out of a sense of curiosity. I bet you use AI like this all the time too.
You have a rough idea of your goals and turn to Claude for clarity. Claude throws some ideas around and suddenly things become clear. Nope, that’s not how it works. What really happens is Claude interprets your goals according to its own understanding. You’re no longer wanting what you want – you want what Claude thinks you should.
Knowing a bit about AI, you know the bar isn’t high when it comes to meeting your needs perfectly. Tomorrow’s no different.
The skill matters because it precedes all others and something you’d miss without this tech. If you truly knew what you wanted, you wouldn’t need this article in the first place.
What happens if you open Claude and ask a question without having a clear picture of what you want? That’s how ‘knowing what you want’ looks in practice. The AI will always respond because it’s designed to answer. It’s not talking to you like a person would; the AI won’t ask for your preferences upfront.
Once you get an answer, one of two things happen. If the AI gets it wrong, you don’t know that. You can’t compare its response to your actual goals because you didn’t set them clearly in the first place. If the AI gets it right, you also won’t realise it and leave with the impression that 1) you’ve mastered using AI and 2) you know what you want. But both are off: you’re relying on an unreliable machine.
If instead you follow my advice and go in knowing your goals clearly, everything changes. When Claude asks for more context, you can provide it easily. If the response is nonsensical, engage with the model until its understanding aligns with yours. After testing against your set objectives, a good result emerges.
Get it? The power of AI depends entirely on your own ability to know what you want.
The most important skill in dealing with AI turns out also being one of life’s most fundamental – but rarely asked questions. That’s not my idea; it stems from 3,000 years of Western philosophy.
Socrates based his whole project on the premise that you don’t know what you want and that’s where all your problems lie. Viktor Frankl found those who survived concentration camps knew why they wanted to live. Kierkegaard argued people adopt preferences unconsciously from their environment. And so on...
AI is today’s version of letting someone else make decisions for you – don’t be that person.
As the manager of AI tools, you can embrace this agency era – setting goals, constraints, tools etc., and having them executed. But it also brings challenges to control like realising your poor ‘what’ skills. You overcome this by learning how to manage well. Realising you don’t know how to do things anymore? Learn how to execute effectively again. The third challenge is about being human – why even do anything at all? What do you want from life? Master these questions, and AI becomes your best colleague.
Don’t let AI pin your expectations on its output. That’s the surest way for AI to drag you towards a mediocre version of yourself. Your untapped potential needs you before it can help unlock your full self. Brain before AI > AI before brain. If you don’t take care of yourself, soon enough you’ll settle for mediocrity in outsourced work.
The problem isn’t that AI says wrong things or does bad stuff. It’s that it stops you from striving for the true, good and beautiful in life.