Technology · 2 min read

AI Is Getting Better at Answers. We Still Need Better Questions.

Author
Kenji Watanabe
Published on Sep 12, 2026
AI Is Getting Better at Answers. We Still Need Better Questions.

Better answers change the bottleneck

For much of computing history, the challenge was getting a machine to execute the right instruction.

As AI systems become better at generating text, code, images, and analysis, another problem becomes more important: deciding what should be asked in the first place.

When answers become cheap, judgment about questions becomes more valuable.

The seductive wrong question

A model can produce a polished answer to an unclear prompt. That does not make the prompt good.

Imagine asking:

Write a strategy for increasing growth.

The system can generate paragraphs. But growth of what? For whom? Over what period? Under what constraints?

Without context, fluency creates an illusion of progress.

Questions define the problem

Good questions reveal assumptions.

Instead of "How do we increase sales?" a better question might be "Which customer segment is most likely to expand without increasing support costs?"

The second question contains a strategic choice.

What humans should still own

AI can help with brainstorming, synthesis, drafting, and pattern recognition.

People still need to decide:

  • What matters.
  • What evidence is trustworthy.
  • Which tradeoffs are acceptable.
  • What outcome is worth pursuing.

Those are not merely input fields for a model.

They are acts of judgment.

The new literacy

Prompting is sometimes presented as a technical skill, but a deeper skill is problem formulation.

A person who understands the real problem can use many tools effectively. A person who does not may produce impressive work in the wrong direction.

This is not a temporary transition.

Every powerful tool eventually shifts the scarce resource somewhere else.

When computation became cheap, data became valuable. When information became searchable, attention became scarce. As generation becomes cheaper, judgment may become the bottleneck.

The future may belong to people who know not only how to ask AI for an answer, but when the answer should change the question.

Author

Kenji is a technology writer interested in AI, software craftsmanship, and the changing relationship between people and machines.

View full profile & stories

Discussion (0)

Have a question or insight to share on this story?

Sign in to leave a comment
Be the first to share your thoughts on this story.