D&AD costed the shortcut. Agencies will find out what judgement was worth in about five years.

D&AD published its AI & Creativity Report 2026 on 18 August. The subtitle is The Cost of the Shortcut, and the finding underneath it is not about software. Entry-level roles are disappearing at the same moment AI absorbs the repetitive work those roles used to consist of. The jobs where judgement got learned are being switched off, and the report is blunt about what a shortcut produces without it: average output.

It is built on 197 conversations with creative leaders across 30 countries and an analysis of more than 10,000 D&AD Award entries. That is a serious piece of work arriving at a conclusion the industry has been walking around for two years.

The same week made the point three more times. Maisie McCabe asked in Campaign whether originality is still a moat, and landed on the view that the proliferation of familiar work might raise the value of uniqueness rather than lower it. Liz Wolstenholme wrote that old ideas, resurrected together, start to look like stale thinking - and she helped make one of the originals. In The Drum, Rob Pratt asked why every AI brand looks exactly the same.

Four writers, one week, one argument. The cheap stuff got cheaper, and the scarce thing is now the call somebody makes about it.

Where the deficit shows up first

Not in the client work.

Client work is the most heavily supervised output an agency produces. It is briefed, reviewed, pitched, argued over, and it eventually goes in front of somebody who is paying for it and paying attention. If judgement is thin, three people catch it before it ships.

A judgement deficit shows up first in what the agency says about itself, because that is the work nobody senior checks.

The release goes out because the release always goes out. The LinkedIn post is somebody's Tuesday task. The award entry gets written in deadline week. The website copy was signed off two years ago by a person who has since left. None of it passes an editor, a client or a jury before it goes.

That material is the first thing a trade journalist sees. It is also what an AI assistant reads when it assembles an answer about your category. The least supervised writing in the building is doing the most exposed work.

The apprenticeship was the mechanism

Everyone who is good at this learned it by doing the unglamorous version badly, in public, with somebody senior explaining afterwards why it did not land.

You wrote the release that got no reply. You sat on the call where the point read completely differently out loud than it had on the page. You watched an editor kill a story you were proud of and, eventually, understood the reason. Attempt, failure, correction, repeat. That sequence is what judgement is made of. It is not a talent and it does not arrive fully formed. It is compressed experience, and the compression takes years of small, low-stakes failures that nobody remembers.

D&AD's proposal is not to preserve junior roles as they currently exist, which is the right call. It is to redesign them so they still teach the thing. That is harder than it sounds, because the task and the teaching were the same object. Remove the task and the lesson has nowhere to live.

The commercial read

If the industry is losing its ability to tell good from average, then being visibly good at telling good from average becomes a position rather than a hygiene factor.

That is the opportunity buried in McCabe's argument, and it is an uncomfortable one. As the volume of competent, familiar, machine-assisted work rises, the return on being distinctive rises with it. Scarcity does the pricing.

For an agency, the practical consequence is that the public record has to carry evidence of judgement rather than evidence of activity. A view that could turn out to be wrong. A specific claim about your market that a competitor would want to argue with. An opinion with a name attached to it and a person prepared to defend it in a room.

Most agency communication carries neither. It carries news - we won, we hired, we launched - which proves the agency exists and settles nothing else.

The five-year problem is structural and nobody is fixing it this quarter. The nearer problem is available now, and it is more embarrassing: in most agencies the people who can actually make the call are not writing anything, and the people writing everything are not being taught to make the call.

If you want a read on what your own public record currently demonstrates, we can start there.