Most arguments about AI and design work are about whether the work survives. That is the wrong question, and it is the wrong question in an interesting way.

The work survives. What is being removed is a specific layer of it: the mechanical half. Resizing a layout to four breakpoints. Producing the eight variants nobody will pick. Writing the boilerplate CSS. Cutting the first draft that exists mainly so someone senior has something to react to.

That work was genuinely tedious and few people will miss doing it. But it was not only tedious. It was the ladder. It was how someone with taste but no experience earned a seat in the room, and how they built the pattern library in their head that later becomes judgement.

Take away the bottom rung and the ladder does not get shorter. It gets unreachable.

What actually gets faster

Worth being specific, because the honest version of this argument does not require exaggerating the tool.

What genuinely accelerates is anything with a known answer and a tedious execution. Generating the twelve states of a component you have already designed. Producing responsive variants from an established system. Writing the utility CSS. Converting a design decision that is already made into markup. First-pass copy for a layout you will rewrite anyway.

The common factor is that the thinking is done and what remains is the typing. That is exactly the category of work that used to be handed down.

What does not

Also worth being specific, because this is where the bull case lives.

Deciding what the thing should be at all. Knowing that a client asking for a homepage redesign has a navigation problem. Understanding why the third option is worse even though it tests better. Holding a design system in your head well enough to know that a proposed component is really two components, or is really an existing one with a modifier.

None of that has got meaningfully faster, because none of it is execution. It is the part of the job that requires knowing why rather than what, and it is the part that was always undervalued relative to how hard it is.

It is also the part that keeps moving. Knowing which of this year’s platform features makes a component simpler, and which just makes it different, is judgement rather than knowledge, and no tool arrives holding that opinion for you.

So the bull case is real: more of the job becomes the interesting part. Anyone who already has the judgement is straightforwardly better off. We ship design systems and the mechanical acceleration is genuinely welcome.

The problem nobody has a good answer to

Here is the part the optimistic version skips.

Judgement is not taught. It is accumulated, mostly by doing a large volume of unglamorous work under someone who corrects you. You learn why a component is really two components by building the wrong version, shipping it, and maintaining it for a year. You learn to see a navigation problem by producing forty homepages.

That volume of work is precisely what has been automated. Which produces an awkward sequence:

  • The junior role existed largely to do the mechanical work.
  • The mechanical work is now cheap.
  • So the junior role is harder to justify commercially.
  • But the mechanical work was how people became senior.
  • And senior people are what the remaining work requires more of, not less.

Every step is individually rational. The outcome is a profession that consumes senior judgement faster than it produces it.

This is not a prediction about ten years out. It is visible now in how studios hire: teams that would once have taken two juniors and a mid take one senior and a subscription.

The second problem: clients who cannot tell

There is a related pressure that gets less attention.

A large share of clients cannot reliably distinguish good design from adequate design. They could not before either. What has changed is that adequate is now available instantly and nearly free.

When the gap between adequate and good was several thousand pounds and three weeks, the decision had friction in it, and the friction was often enough to get a proper conversation started. When adequate costs nothing and arrives in a minute, a client has to actively choose to pay for a difference they cannot see.

Some will, because they have been burned, or because the stakes are obvious, or because they have taste themselves. Many will not. That is not a moral failing on their part; it is a rational response to an option that did not exist before.

The work that survives this is the work where the difference is legible: where the design has to hold up under real content at real scale, or integrate with a system, or be maintained by someone else in two years. Where good and adequate diverge visibly and expensively.

What a studio can actually do

Not much of this is solvable at the level of an individual team, but some of it is, and it is worth doing deliberately rather than waiting to see what happens.

Give juniors the judgement work earlier than feels comfortable. The traditional sequence was years of execution followed by decisions. That sequence no longer works, because the execution years have been compressed out of existence. Put people in front of the actual decision sooner, with more supervision than you would have given before. It is slower per project and it is the only version that produces someone senior at the end.

Make review the teaching, not the gate. When the mechanical work was done by hand, review could be light, because the doing taught most of the lesson. Now the doing teaches less, so the review has to carry the weight. That means explaining why rather than marking up what, and it means senior time budgeted as teaching rather than borrowed from delivery.

Pair on the ambiguous work specifically. Not on the parts a tool would do anyway. On the parts where someone has to decide, and where a junior watching the decision get made is the whole point.

Be honest in hiring. A junior role that is now mostly prompting and checking output is a different job from what the title used to mean, and describing it accurately is better than the alternative for everyone.

Where both sides agree

Strip away the framing and the optimistic and pessimistic readings converge on one point: the work that survives is the work that requires knowing why.

The optimist says this is good, because that was always the valuable part and now it is what the job consists of. The pessimist says the route to being able to do it has been removed. Both are correct, and they are not actually in tension. It is entirely possible for the remaining work to be better and for fewer people to be able to reach it.

Which makes the interesting question not whether design work survives, but who gets to do it in ten years, and whether that group is larger or smaller than today.

The honest answer is that on the current trajectory it is smaller, and that nothing about that is inevitable. It is a consequence of a lot of individually sensible hiring decisions, and it can be changed by making different ones on purpose.

Frequently asked questions

Is this an argument against using AI in design work?

No. The mechanical work was tedious and automating it is a straightforward improvement to the day. The argument is that removing it has a second-order effect on how people enter the profession, and that effect needs handling deliberately rather than assuming it works itself out.

Have not previous tools caused the same panic?

Yes, and mostly the panic was overdone. Desktop publishing, stock photography, page builders and website builders all removed layers of work and the profession absorbed each one. The difference worth watching is that those tools removed work at the bottom and left the ladder intact, because there was always a next tier of mechanical work to learn on. This time the mechanical tier itself is what went.

Should a junior designer be worried?

Worried is not useful. What is useful is skipping straight to the judgement work rather than expecting to earn it through years of execution, because those years are not on offer. Take the decisions, get them wrong under supervision, and build the pattern library faster than the previous generation had to.

Does this apply to development as well as design?

The same shape, yes. The mechanical tier of front-end work has been compressed in the same way and for the same reasons, and the route from junior to senior ran through the same kind of volume. The specifics differ; the structural problem does not.

What about designers who never had the mechanical phase?

They exist and some of them are excellent, which is the strongest argument that the ladder is not the only route. But they are usually people who built the pattern library some other way, often through volume of personal work. The route matters less than the volume, and volume is what has become optional.

The bottom line

AI removed the mechanical half of design work first, because that half had a known answer and a tedious execution. It was also the half juniors did, and doing it was how they accumulated the judgement that the remaining work requires more of.

That is not an argument for doing the tedious work by hand out of nostalgia. It is an argument for noticing that the training ladder was load-bearing, and that if you want people who can do the surviving work in ten years, somebody has to deliberately build a replacement for the rung that just disappeared.