We were told designers would be replaced. Two and a half years on, that hasn’t happened. What has happened is stranger and, I think, harder to navigate than replacement would have been.
The floor came up
Anyone can now produce something that looks competent. A reasonable layout, a defensible type pairing, a passable illustration, a serviceable paragraph of marketing copy. The baseline of visual acceptability has effectively become free.
This is genuinely good. Small businesses that could never afford a designer can now put something decent into the world. A founder can get to a credible-looking first version without a budget. I don’t think there is much to mourn there.
And so did the flood
But because competent is free, competent stopped being worth anything. The same tools that raised the floor demolished the barrier to mediocrity. The volume of plausible, forgettable work has gone up enormously, and it all sits in the same feeds and inboxes as everything else.
The practical consequence for anyone doing this professionally: the work that used to distinguish you — clean, on-brief, well-executed — now describes the baseline that anyone can hit in an afternoon.
What became more valuable
Judgment. Knowing which problem is actually worth solving, which is not something you can prompt for, because it requires having sat with the client, the constraints and the politics.
Taste, in the specific sense of being able to look at twenty generated options and know which one is right and why. The generation is free; the selection is the job.
The ability to say no. To a brief that is wrong, to a direction that is fashionable and hollow, to the fourth round of changes that will make the work worse. Models are agreeable by construction.
And context. A model does not know that this client’s last campaign failed for reasons nobody wrote down, that the CEO hates the colour, or that the audience in this market reads that gesture differently. That knowledge is unevenly distributed and largely unwritten, which is exactly why it holds value.
What I actually use it for
First drafts I fully intend to discard. Getting something ugly onto the page to react against is faster than starting clean, and reacting is easier than originating.
Breadth early on. Twenty directions in an hour, most of them bad, occasionally one that shifts how I’m thinking about the problem.
The unglamorous writing. Alt text, meta descriptions, the third variation of a form error. Necessary, and not where my attention is best spent.
Arguing with my own reasoning. Asking for the strongest case against a direction I like is genuinely useful, if you can resist dismissing the answer.
What I don’t use it for is the last stretch of craft, or anything requiring me to know something about a specific human being.
It is good at the middle and bad at both ends
This is the pattern I keep noticing. Generative tools are strong through the middle of a project — exploration, iteration, production, variation — and weak at the two ends.
They cannot do the beginning, because the beginning is discovery, and they have never met your users, walked the site, or sat in the room where somebody said the thing that reframed the brief.
They cannot do the end either. The last five percent — the kerning, the timing of a transition, the decision to remove a whole section — is a series of judgment calls that only make sense against a specific context.
The part nobody has solved
The traditional apprenticeship in this industry ran through production work. You resized banners, prepared files, built out the boring pages, and while doing it you absorbed a thousand small judgments from the people reviewing your output.
That work is exactly what is disappearing first. Which means the ladder juniors climbed is being removed while the expectation that they arrive with judgment stays. I don’t have a tidy answer, but I think teams should be deliberate about this rather than discovering the gap in five years.
The job was never pushing pixels. It was deciding what is worth making, and then making it properly. The first half of that has become more valuable, not less. The second half now has a much higher bar for what counts as properly.