For a while, the dominant story about software was that it would remove judgment from work. We would encode the rules, automate the obvious choices, and leave fewer decisions exposed to the unreliability of people.
That story was never entirely wrong. It was simply incomplete.
When information was expensive, knowing was a source of leverage. You needed access to the manual, the archive, the experienced operator, or the room where the decision had already been made once. Better software made more of that knowledge available. Artificial intelligence has accelerated the trend dramatically.
The result is not a world without judgment. It is a world in which judgment becomes easier to see.
Answers are becoming cheap
Most professional environments contain an enormous layer of questions that are difficult only because the information required to answer them is fragmented. A good model can now reconstruct a plausible answer from documentation, examples, and patterns in a fraction of the time.
That changes the bottleneck. The hard part is less often producing an answer. It is deciding which question deserves attention, which answer fits the actual situation, and which second-order effects matter.
This is why good operators can appear almost unnervingly calm. They are not processing every available signal. They are selecting the small number of signals that change the decision.
Judgment is an operating system
Judgment is sometimes treated as taste with better manners. In practice, it is closer to a compact operating system built from experience:
- a model of what tends to go wrong;
- a sense of which constraints are real and which are inherited;
- an ability to notice when a familiar situation is not actually the same;
- and the courage to make a choice before every ambiguity has disappeared.
Tools can support each of these capabilities. They can also produce enough plausible output to overwhelm them.
The people who benefit most from powerful tools will not merely be the people who can generate the most. They will be the people who can reject the most intelligently.
The return is quiet
There will be no ceremony marking the return of judgment. It will show up in small differences: the product leader who recognizes the exception that breaks the workflow, the underwriter who understands which missing fact matters, the editor who removes the paragraph that sounds impressive but weakens the argument.
Automation does not end discretion. It moves discretion to a higher layer.
That is good news for anyone willing to practice it.