There's a mildly entertaining turf war happening in the design community at the moment, and neither side are getting it right.
On one side, you've got AI companies and their investors insisting that design is effectively dead. AI can generate layouts, build design systems and write the CSS to bring it to life - all from a single prompt. Why would you pay a designer £50k a year when a machine can do the same thing in seconds for less than a cup of coffee? On the other side, you've got designers desperately clinging to the idea that AI will never replace them because it can't create - it can only aggregate. It doesn't understand emotion. It doesn't have taste. It can't make anything genuinely original.
Both positions are partially right and mostly self-serving. And neither side seems willing to engage with the more interesting question - which is not whether AI replaces designers, but when each one is more appropriate than the other.
Follow the money
Before we take anyone's predictions too seriously, it's worth remembering who's making them.
The frontier AI labs - OpenAI, Anthropic, Google DeepMind et al - are burning through money at a staggering rate. OpenAI posted a $13 billion net loss in the first half of 2025 alone and is projected to lose $14 billion in 2026. Cumulative losses between 2023 and 2028 are expected to hit $44 billion before the company turns a profit - if it ever does.
HSBC has estimated that OpenAI faces a $207 billion funding shortfall to sustain its growth plans. These are not companies operating from a position of quiet confidence. They need founders and business leaders to believe that AI can replace their workforce, because without that belief, the investment dries up - and the investment is the only thing keeping the lights on.
So when an AI CEO tells you that design is dead, remember that he's not giving you an objective assessment. He's selling you a dream.
Designers, meanwhile, have their own incentives for dismissing AI altogether. Nobody wants to hear that their craft is being commoditised. So they retreat to the high ground of creativity and human emotion - "AI can't create, it can only copy" - and convince themselves that this makes them safe.
There is some truth in that, but it's not the whole truth, and it conveniently ignores the fact that a significant chunk of what most designers actually do on a day-to-day basis is copying - assembling established patterns, reusing components, following conventions that have been well understood for years.
The MAYA principle
There's a concept from the mid-twentieth century that explains what's really going on here far better than either side is managing.
Raymond Loewy - the industrial designer behind the Coca-Cola bottle, the Shell logo and even the nose of Air Force One - developed what he called the MAYA principle: Most Advanced Yet Acceptable. His theory was that humans are caught between two opposing forces - a love of new things and a fear of anything too new. The sweet spot is where something feels fresh and exciting without being so unfamiliar that it frightens people off. The writer Derek Thompson distilled it quite neatly: "To sell something surprising, make it familiar. To sell something familiar, make it surprising."
The iPhone is probably the most famous example of this in action. It was essentially a pocket computer - a genuinely radical piece of technology - but Apple called it a phone. The industrial design echoed the iPod that everyone already owned, and it kept familiar functions like making calls and sending text messages front and centre. The concept was revolutionary, but the packaging was deliberately conventional. People didn't have to adapt to a new category of device - they just had to accept that their phone could now do more.
ChatGPT is another example. The underlying technology - large language models - had been available via OpenAI's API for years, but nobody outside of the developer community even knew about it. The breakthrough wasn't the model itself - it was the decision to wrap it in a chat interface. Everyone already understood chat from WhatsApp and SMS. That single, familiar design decision made genuinely alien technology feel as natural as texting a mate - and arguably kicked off the entire AI boom. It's worth noting that the most important design decision in the history of AI wasn't made by AI. It was made by a human who understood their audience.
How does this relate to the looming AI vs Designer death-match?
When AI runs out of steam
MVPs don't stay MVPs forever. If the idea proves itself and starts gaining traction, something changes. Competitors notice. They build their own versions. And because most of them are using the same AI tools, trained on the same data, following the same patterns - they all end up looking remarkably similar. You've gone from having a distinctive new idea in familiar packaging to having a familiar idea in identical packaging. At that point, you're in a commodity market, and that's where AI design falls apart.
This is where human designers earn their keep. When your product has become the new normal and the market is full of derivative copies, you need the packaging to be bold, surprising and memorable. You need the opposite of what AI gives you. You need someone who can look at the conventions everyone else is following and deliberately break them in ways that feel exciting rather than confusing. That takes taste. It takes an understanding of your audience that goes beyond data and patterns. It takes the kind of creative judgement that comes from years of practice and an intuition for what feels right, not just what looks right.
AI can't do this. Not because of some romantic notion about human creativity being sacred, but because of how large language models and image generators actually work. They produce output that converges towards the mean - the average of everything they've ingested. They're brilliant at reproducing what's already common, and structurally incapable of producing something that breaks away from it. You can't generate a bold, original design by averaging together every design that already exists. That's not a limitation that will be fixed with better models; it's a feature of the architecture.
The 80/20
In my experience - and this is gut feel from twenty-odd years of designing software for a multitude of industries, rather than a scientifically validated ratio - about 80% of the design work in any product is conventional. Established patterns, familiar layouts, standardised interactions that users expect and that don't need reinventing every time. Navigation, forms, tables, settings pages, dashboards. This is the stuff AI can handle right now, and will only get better at.
The remaining 20% is the work that actually defines your product - the signature experiences, the moments of delight, the interactions that make someone say "this is better than everything else." That last 20% requires a designer who understands the audience on a human level, who can make creative decisions that AI would never arrive at, and who has the intuition to know when a convention should be followed and when it should be broken.
The designers who will thrive are the ones who can work at both levels - using AI to handle the 80% at speed, while getting hands-on for the 20% that actually matters. They'll write better prompts than non-designers, because they know what good looks like and can articulate what they want. And they'll know when to stop prompting and when to start designing, because they can feel when the AI's output has hit a point of diminishing return.
The designers who are in trouble are the ones who were already producing work that was indistinguishable from what AI generates - following patterns, assembling components, never deviating from the design system. If your entire job was to produce averaged work, then yes, AI is coming for you. Not because AI is brilliant, but because your work was already average.
Neither side wins
The AI companies are wrong - design isn't dead, and replacing your entire design team with a prompt is a great way to build a product that looks exactly like everyone else's. But the designers are wrong too - pretending that AI has no role in design because it lacks human creativity is wilful blindness to the 80% of the job that doesn't require it.
The answer, as is usually the case, is way more boring than either side wants to admit. AI handles the conventional work, faster and cheaper than humans ever could. Designers handle the creative work, the stuff that requires judgement and taste and an understanding of people that no model can approximate. The skill, for product leaders and design teams alike, is knowing which is which - and having the discipline not to use one where you need the other.
