Every conversation about AI and design eventually collapses into one of two takes.
Take one: AI is going to replace designers. Figma plugins can mock up dashboards from text prompts, and that's the end of the profession.
Take two: AI is just a tool. Nothing fundamental changes. We've had this conversation before with Photoshop, with Sketch, with Figma.
Both are wrong. More importantly, both are boring.
The interesting question is narrower. What specifically does AI change in the shape of the work? What doesn't it change at all?
What it changes
AI is very good at the middle of the design process. The part with high volume, low variance.
Generating ten variants of a layout to explore. Rewriting twelve microcopy strings to fit a 32-character limit. Transcribing and tagging a research interview. Checking a color palette against WCAG across every state. Producing three alternative empty states before you've decided on any of them.
A designer who uses these tools well moves through the tedious middle faster. The question is where that saved time goes.
What it doesn't change
AI is bad at the beginning and the end.
The beginning is knowing what problem you're solving and why. An LLM can't tell you whether the feature your PM is asking for actually matters to your users. It can draft three onboarding flows in forty seconds, but it can't tell you whether users need onboarding at all. You have to go find out. Which means talking to people. Which hasn't gotten faster.
The end is taste, trust, and context. Is this the right decision for this team, this product, this market, this user? No probabilistic model knows that. It has context on everything in its training set and no context on your specific situation. You're the only one who does.
The parts AI accelerates are the parts that needed the least care to begin with. The parts it doesn't accelerate are the ones that matter, and that's no coincidence: care takes attention, and attention doesn't scale.
What the work is for
Most of the discourse misses this part. AI changes what designers do, yes, but it also changes what the work is for.
Before, a lot of our value as designers came from producing design artifacts. Wireframes, prototypes, mockups, specs. Production was hard and slow. A designer who could produce a lot of artifacts, quickly, was valuable.
Now production is cheap. A reasonably prompted model generates more artifacts in an afternoon than a designer used to make in a week. The artifacts themselves are no longer where the value lives.
The value shifts to everything upstream and downstream of production. Upstream: framing, research, critique, decision-making, alignment. Downstream: implementation quality, iteration, measurement, care. The artifact is now the cheapest part of the process. The expensive parts are thinking clearly about the problem and making sure the shipped thing actually works.
Designers who respond to AI by trying to produce more artifacts faster are running the wrong race. The better response is to get deeper at the parts AI can't help with.
What to do next
Three concrete moves if you're a designer reading this.
Use the tools. Seriously use them. Don't have a take without hands-on experience. Use a coding assistant. Use Figma's AI features. Use Claude or ChatGPT to rewrite copy. Whether you like it or not, some of the conversations you'll have next year will assume your counterparts have used these. Be informed.
Move your effort upstream. If AI saves you two hours producing variants, spend those two hours talking to users. Or reading product usage data. Or arguing with your PM about priority. The tools are a gift to the parts of the job that used to get squeezed for time. Don't waste the gift by generating more variants you didn't need.
Pick what you refuse to automate. Some parts of the work are worth protecting on principle, whether or not AI could do them, because doing them yourself is the point. Qualitative interviews. Design reviews where you actually sit with the work. The first sketch of a new problem. Decide where your hands stay on the material, and hold that line.
Where we stand
We use these tools, we recommend them, and the teams we work with use them.
Our problem is that the discourse has collapsed into either "AI eats design" or "AI is nothing new," and neither frame helps you make decisions on Monday morning. What helps is knowing what actually changed, what didn't, and which parts of your practice are worth protecting even when you don't have to.
Design was never primarily about production. We could pretend otherwise for a while, because production took long enough to feel substantial. It doesn't anymore, and what's left is what the job always was: judgment, care, and attention to people.
This is one of the arguments at the core of the Philosophical Foundations workshop: how to think clearly about product decisions when the tooling is accelerating faster than the frameworks around it.
