Good designsurvives the toolthat made it.

Does AI make design judgment less important?

No. AI and other generative tools reduce the effort required to produce plausible options. That makes judgment more important because the volume of material grows faster than the time available to examine it.

A new production method can change the route to an image, interface or system. It does not change the questions that determine quality: who is this for, what must it help them do, what evidence supports the choice, what should remain consistent and what can be removed?

Tools change faster than principles

Useful design principles endure because they describe relationships between people, tasks and systems rather than features of a particular application. GOV.UK's principles begin with user needs, doing less, designing with data and iterating. Those ideas remain valuable regardless of the software used to express them.

This does not mean tools are irrelevant. Each tool changes what is easy, what is visible and which mistakes become common. The designer's responsibility is to understand those conditions without mistaking them for the purpose of the work.

Generating is not selecting

Generation widens the field. Selection gives the field direction. The second action depends on criteria: relevance to the brief, distinction, clarity, accessibility, feasibility and coherence with the larger system.

Without criteria, more options create aesthetic drift. Teams begin reacting to individual moments rather than building a consistent whole. Selection is choosing the direction that can carry the most responsibility, not the frame with the quickest impact.

Generating is an action. Selecting is a responsibility.

Speed makes direction more important

When a new option takes minutes rather than days, it becomes easy to confuse activity with progress. Progress occurs when an iteration answers a question, removes uncertainty or strengthens the system.

The Double Diamond separates exploration from definition and development from delivery. The model remains useful because divergence only creates value when convergence follows. Faster exploration should create more room for evaluation, not an excuse to skip it.

Systems outlast screens

A screen can be convincing in isolation and weak in a product. Design systems make relationships explicit: typography, spacing, states, components and behaviour. They allow the work to remain coherent as content changes and routes appear.

The Decimal Nine interface improved when repeated values became shared alignment contracts rather than page-specific nudges. The visible result is not the token itself. It is the consistency that survives across services, viewports and later revisions.

Novelty is not a design objective

New tools encourage a recognisable aesthetic because their defaults spread quickly. That can be useful during exploration, but a default is not a point of view. Design becomes specific when it responds to the organisation, audience, content and task rather than displaying the signature of the software.

The test is not whether the work looks as though AI was used. The test is whether the result communicates the right thing with enough distinction and restraint. A production method can remain invisible when the design itself carries a clear reason for being.

This is why authorship still matters. Authorship is not manual control over every pixel. It is responsibility for the relationships that remain: what leads, what recedes, what repeats, what changes and what the work refuses to include.

Consistency needs room for judgment

A design system should not turn every page into the same template. It should define the stable decisions that let meaningful variation occur: gutters, type roles, spacing rhythms, interaction states, accessibility behaviour and responsive principles.

The Thinking publication uses one mathematical grid while giving each article a different typographic expression. The distinction is important. Consistency lives in origins, measures and behaviour; character lives in how those constraints are composed.

Tools can accelerate the creation of variants, but the system decides which differences are useful. A variation earns its place when it communicates meaning, not merely because it is possible.

Tool literacy includes knowing what to ignore

Fluency is often measured by the number of features a practitioner can use. More important is the ability to recognise which features do not belong in the work. Every tool encourages particular effects, defaults and workflows; unexamined fluency can make those defaults look inevitable.

A strong practice uses the tool selectively. It may use automation to organise material, generation to test range and code to verify a responsive system. It may reject an impressive effect because it delays access to content or weakens the hierarchy. Restraint is not a refusal of capability. It is capability directed by purpose.

The standard remains outside the software. The work should earn attention, communicate clearly and operate under real conditions. A useful tool helps reach that standard. It does not define it.

What should survive the tool

The problem definition should survive. The hierarchy should survive. The content should remain understandable without motion. The interaction should work with a keyboard and on a narrow screen. The implementation should be maintainable by someone who did not make the first composition.

These are not constraints that diminish creativity. They are how creative intent remains available after the original file, tool or operator has moved on.

The Decimal Nine test

We ask whether the design can explain its decisions without referring to the novelty of the tool. Does the hierarchy guide attention? Does the system remain clear under real content? Does the composition re-form deliberately on mobile? Can the interaction be reduced without losing meaning? Can the result be built and maintained?

If a design only feels intelligent because the tool used to create it is new, the design itself is probably not doing enough. Good design survives the tool because its intelligence lives in the relationships it creates.