Judgement Is the Thing You Cannot Automate

The last essay ended with a line I almost cut. “Choosing is the hard part now.” It sounded too clean. But I kept it because it was true, and because I did not yet know how to say the harder thing behind it.

Here is the harder thing. Choosing is not a skill. It is not something that gets better because your tools got better. If anything, it gets worse.

I spent the last few months building things. A fiction generator for Marathi literary writing. A brand system for a story platform. A game prototype with my son. This essay series. I built all of these in roughly the same period, sometimes switching between them in the same afternoon. The tools made this possible. What the tools did not do is tell me which of these deserved my morning, which deserved my evening, and which deserved to be killed.

That is judgement. Not intelligence. Not strategy. Judgement.

I know smart people who build the wrong things with extraordinary skill. They can reason through any problem, break down complexity, ship fast. But they cannot tell the difference between a project that is alive and a project that is merely working. These are not the same thing. A project can have users, generate revenue, and still be dead. You feel it in your chest before you see it in the numbers.

Judgement is what registers that feeling and acts on it.

Most people think judgement is about saying yes to the right thing. It is not. It is about saying no to good things. That is what makes it expensive. Saying no to a bad idea costs nothing. The expensive no is the one you say to an idea that works. People want it. You could grow it. And you say no anyway, because it is not yours, or the timing is wrong, or something else is more alive. That costs something real. It feels like waste. Sometimes it feels like grief.

I do not think you can learn judgement from a book or a course. You cannot framework your way into it. It accumulates. You build something and it fails. You build something else and it works but feels hollow. You watch other people build things that look impressive and then quietly disappear. Over time, these experiences leave a residue. Not rules. Something closer to instinct. You start recognizing patterns you cannot fully explain. This idea has weight. That one does not. You do not always know why. But you trust the signal more than you trust the argument.

This is also why judgement is scarce. It does not transfer well. My judgement is trained on my failures, my context, my specific history of building things in Marathi digital content and AI products and consulting. Your judgement is trained on yours. We can share frameworks, but frameworks are not judgement. They are the scaffolding people build when they do not yet have judgement of their own.

Now here is the part that surprised me.

I use AI as a sparring partner. Not to make decisions, but to pressure-test them. I describe an architecture and ask it to find the weaknesses. I lay out three options and ask it to argue for the one I like least. This works better than it has any right to. AI is a better sparring partner than most humans for this, because it has no stake in the outcome and no ego in the argument.

But it works only because I know what to ask. The quality of the sparring depends entirely on the judgement I bring to it. When I ask a sharp question, I get a sharp response. When I ask a vague question, I get a confident, well-structured, completely useless answer.

I have taken this further. In the fiction generator, I built a system where multiple agents argue with each other. One generates, one critiques, one checks for cliche, one evaluates emotional honesty. A council that fights. The idea was that judgement could emerge from disagreement. And in some ways it does. The output is better when the agents are in tension than when any single agent runs alone.

But here is what I learned from building it. The council does not have judgement. I have judgement. The council runs on the judgement I encoded into it. Which agents I included, what I told the critic to care about, what I defined as a valid objection. Every design choice was a judgement call. The system did not replace my judgement. It amplified it. It ran my taste at scale.

People assume AI removes the need for human judgement. My experience is the opposite. AI is a multiplier. Good judgement in, better output. Mediocre judgement in, confidently mediocre output at scale. The tool does not correct for the quality of the thinking behind it.

And it gets worse. When execution was hard, you could hide behind constraints. I would have made the right call if I had more time. I would have built the right thing if I had better tools. The constraints were real, but they were also cover. They gave you a reason for not being more decisive.

AI stripped that cover away. Now you have the time. You have the tools. You have the ability to try anything. And every bad call is yours. There is no friction left to blame.

More capability means more options. More options means more decisions. More decisions means more judgement required. The person who can build ten things in the time it used to take to build one does not need less judgement. They need ten times more of it.

So judgement is not getting automated. It is getting more expensive. The cost is not money or time. It is the willingness to kill your own ideas, to sit with uncertainty, to choose before you have complete information, and to live with the fact that you might be wrong.

I have been thinking about whether there is a layer beyond judgement. Something finer. Judgement tells you what to cut. It is a filtering function. I am living this right now. The projects I listed at the start of this essay are the ones that survived my own filter, and the filter cannot rank them. Each one is defensible. Each one is viable. Judgement got them here. Judgement cannot tell me which of them deserves the morning.

Because the difference between the remaining three is not that one is right and two are wrong. The difference is that one fits and two do not. And “fits” is not something you can argue for. It is something you recognize.

That recognition is not judgement anymore. It is something else.

The word I keep coming back to is taste.

Judgement tells you what to cut. Taste tells you which of the survivors deserves to exist. It is harder to build, easier to lose, and so far it is the only layer the machines have not touched.