The part of the job that does not get automated

Nielsen Norman Group's State of UX 2026 names what survives automation as curated taste, research-informed contextual understanding, critical thinking and careful judgment.

The list is accurate as a diagnosis and useless as a job description. Nobody can hire against curated taste, and nobody can teach it from a bullet point. The same report says that a designer assembling components from a design system is already replaceable, which is a harsh reading until you look at what enterprise design work consists of on an average Tuesday.

So it is worth defining the residue concretely enough to do something with.

Knowing which option the situation calls for

Generative tools are good at producing options and bad at choosing between them. The choosing depends on things that are not in the prompt: which of two user groups is strategically more important this year, which constraint the engineering team has already spent its budget on, which of the three plausible patterns the rest of the product has taught people to expect.

That knowledge is contextual, it decays, and it lives in people who have been in the room. It is also the single most teachable part of the list, because the reasoning behind a real decision can be written down.

Knowing what a claim rests on

The second component is the habit of asking where a statement came from. A readout says users find the approval flow confusing. On what evidence, from how many people, doing which task, compared with what.

This matters more now than it did, because fluent unsourced assertions are cheap to produce. A designer who reflexively asks for the provenance of a finding is doing the work that stops a product being built on a plausible sentence.

Generative tools are good at producing options and bad at choosing between them. The choosing is the job.

BrilliantUX editorial principle

Knowing what a decision will cost later

The third is the ability to price a decision over time. A pattern that saves a week now and forks the design system costs more than it saved. A shortcut in the information architecture becomes a migration.

This is the part that only comes from having lived through the consequence of an earlier decision, which is why it is hard to hire for and why it is worth documenting when it happens.

Making it teachable

All three become teachable through the same artifact: a written record of real decisions with the reasoning attached. What was assumed, what the constraint was, what the trade-off was, and what happened afterwards.

A team that keeps that record can onboard judgment rather than hoping for it. A team that does not has to rehire it every time someone leaves.