Whoever has no solid craft won't get better through AI. They'll get faster at being mediocre. That's no opinion against AI — it's an observation from practice.
Figma's State of the Designer 2026 delivers interesting numbers on this. Over half of all designers and hiring managers say AI competence is no longer an option. 91% of designers who actively use AI report better results. 89% work faster. That sounds clear. And yet these numbers only tell half the story.
#What the numbers don't say
91% better results — but better compared to what? To their own earlier work. That's a subjective assessment, no objective quality measurement.
If someone previously produced weak designs, they now maybe produce weak designs in half the time with more variants. That's no progress. That's accelerated standstill.
The Figma article lists five skills designers should sharpen. AI fluency is at the top. Prompting, prototyping, vibe coding. That's correct — these skills are needed. But they're the tool, not the foundation.
#The Midjourney problem
I can generate a hundred image variants with Midjourney in ten minutes. That's impressive. And completely worthless if I can't judge which ones work.
Composition, hierarchy, context, audience — those aren't romantic designer concepts. That's the toolkit that tells me: this variant, not that one. Without this judgement I scroll through a hundred options and still pick the wrong one.
AI generates. You evaluate. If you don't know how to recognise a good solution, no amount of output helps you.
#Where AI is really blind
Figma rightly names user understanding as a critical skill in their article. I'd go further: AI is structurally blind to everything that requires context.
It can't conduct user interviews and sense when someone hesitates. It doesn't understand why a certain company can't implement a solution — even if it's technically perfect. It doesn't recognise when a design is culturally wrong.
Systems thinking is another point. AI can generate components. A consistent design system that works across three years and ten product teams — that's human work. That requires decisions that go beyond the current screen.
Critical judgement, systems thinking, user understanding: those aren't skills AI takes over. Those are skills AI makes visible — because without them the output stays worthless.
#The difference between lever and crutch
There are two ways to use AI. As lever or as crutch.
As lever, AI amplifies what you already can do. You know where you want to go — AI helps you get there faster. You evaluate, decide, steer. The output is better because you are better.
As crutch, AI replaces what you haven't learned. You generate because you don't know how to start. You pick because you don't know what's right. You rely on AI output because you can't judge it.
The difference isn't the tool. The difference is you.
#What this concretely means
Here are three things I consider more important than the next prompting tutorial:
Practise judgement actively. Look at other people's designs daily and argue why they work or not. Not by gut feeling — argumentatively. What exactly makes the hierarchy weak? Why doesn't the CTA pull?
Work on real user problems. Not on mockups for portfolio projects nobody uses. Talk to real people about real problems. That's the ability AI can replace least.
Understand the system before you automate. Before you use AI for design decisions: can you explain why the result is right? If not, you're automating ignorance.
AI fluency is important — I don't dispute that. But it's the last puzzle piece, not the first. Whoever first learns the craft and then adds AI has a real lever. Whoever takes AI to circumvent the craft builds on sand.
You set the direction. AI accelerates the execution. That's the right order.
