AI made execution cheap. That is bad news for most designers.
For twenty years the industry sold execution and called it craft. That business is closing. The one replacing it is harder, smaller, and considerably more interesting.
Kiefer Likens · · 6 min
There is a particular kind of designer who is about to have a very bad decade, and it is not the one everybody expects. It is not the junior. It is the competent mid-career professional whose entire value proposition was being able to produce a clean, on-brief, technically correct deliverable faster than the client could. That was a real skill. It was worth real money. It is now approximately free.
I want to be careful here, because there is a comforting version of this argument that I do not believe. The comforting version says the machines produce generic work, therefore human taste survives, therefore nothing important changes. That is wrong in the way that most reassuring things are wrong: it is true this year and it is load-bearing for an argument that needs to hold for twenty.
The output is getting better. It will keep getting better. Assume it eventually produces work indistinguishable from a good professional’s third draft, and then ask what is left. Because whatever is left is the job.
What is actually scarce
Ask for ten versions of something and you will get ten versions. Several will be fine. One or two will be genuinely good. The bottleneck moves instantly and completely from producing them to knowing which one is which — and then to the considerably harder act of throwing the other nine away without regret.
Everyone can now make the thing. Almost nobody can tell which version is good.
That judgment has always been the actual profession. It was simply hidden inside the production time, where it looked like part of the labour rather than the whole point of it. Strip the labour out and it stands exposed, which is uncomfortable, because a lot of people discover they were being paid for the labour.
And judgment does not come from nowhere. It is accumulated: from having shipped things that failed and understanding why, from having watched a business use what you made, from having been wrong in public often enough to develop a nose for it. None of that is available at speed, which is precisely what makes it durable.
The second thing that survives
The other half is knowing what to make in the first place. A model will answer whatever question you put to it, with total confidence and no opinion about whether the question was the right one. It will cheerfully design a beautiful website for a business whose actual problem is that it is charging half what it should.
Most of the work I get asked to do is described as a design problem and turns out to be a positioning problem wearing a design problem’s clothes. The site is not underperforming because of the site. It is underperforming because the offer is unclear, or the audience is wrong, or the company is quietly embarrassed about its price. No amount of production capacity finds that. Someone has to sit with the owner and ask uncomfortable questions until the real thing surfaces.
What this means if you make things
Get closer to the decision. Not to the pixels, not to the framework, not to whichever tool is being announced this month — to the room where somebody is deciding what the business should do. That is the only position in the chain where speed of production is not the variable being competed on.
And develop taste on purpose rather than by accident. Look at more work than you make. Look at old work, from before any of the current tools existed, because it was made by people who had to be certain before they started. Learn to say why something is bad in specific terms, out loud, to someone who can argue back. That skill compounds and it cannot be prompted.
The uncomfortable summary is that the tools are a gift to people who already had judgment and a serious problem for people who had throughput. I do not think that is fair. I do think it is what is happening.