I’m not a designer. I’ve said this out loud enough times now that it barely stings, but let’s be specific about what it means. I cannot pull together visual materials that look good. My spacing is all over the place. I’ve got no idea what the science behind font weighting is, and if you handed me a blank page and asked me to build something that communicates visually, I’d have no chance. I can tell you when something doesn’t look good. I just can’t tell you how to fix it.
So when designers tell me they’re worried about AI coming for their jobs, I understand the fear, but I think most of them are worried about the wrong thing. The risk isn’t that AI ‘replaces design’ - it’s assuming that ‘design’ is a fixed thing, a specific set of skills and tasks and tastes and worrying that AI will eventually be ‘good’ at them all.
I’ll build a website or a social post, and for a minute I’ll think it looks great. That font feels unique. That colour combination feels like mine. Then I start noticing it everywhere - other founders, other small businesses, all landing on the same fonts, the same palettes, because we’re all prompting the same tool in roughly the same way. AI is doing exactly what it does with language: defaulting to the mean, the middle, the most common combinations. I can tell you when something doesn’t look good. I still can’t tell you why, or how to make it different from everyone else’s.
And herein lies the cause for my optimism… The thing I’m missing isn’t a new tool. It’s the underlying fluency - knowing why a layout works, what a typeface communicates, how spacing changes how something is read. AI doesn’t close that gap for me, because I don’t have the vocabulary to ask it for anything beyond “make it look nicer.” For someone who does have that vocabulary, the same tool stops being a leveller and starts being a multiplier. The designer who understands the language of design can direct AI with precision I’m simply not capable of. That’s not a marginal difference in output quality - it’s the difference between someone who can direct the tool and someone who can only accept what it gives them.
And the places that advantage is needed are multiplying, not shrinking.
We’ve spent the last couple of years mostly using AI through chat interfaces - a box, a prompt, a reply. That’s already old news. AI is moving beyond the chatbot - for example, into voice as the way we interact with software, with digital systems, with our fridge or whatever it might be. Ambient, embedded, device to device. None of that existed as a design problem five years ago. All of it needs someone who understands interaction, not just interface - someone thinking about tone of voice, pacing, how a system should sound when it gets something wrong, as much as how it should look.
Then there’s agents. We’ve always designed websites for human eyes - visual hierarchy, imagery, the stuff that makes a site feel trustworthy at a glance. But a site increasingly built to be read and acted on by an AI agent needs something different: textual structure matters more than visual polish, because the agent isn’t looking at your hero image. (It’s worth looking at what Netlify are predicting around an ‘agent-to-agent internet’, because that shift is already having a huge impact on how sites get built). These products have two audiences now, a human setting the goal and an agent executing it, and designing for only one frustrates both. Nobody has fully worked out what that means for design practice yet.
The third frontier is less visible, but I think it’s the biggest. I work with teams inside large organisations who are being handed tools like Copilot or agent builders and told to go design their own internal software - dashboards, automations, small apps nobody asked a designer to touch. A lot of these people have never had design training, in the same way they might never have had formal engineering training. It’s fine (up to a point) if people are designing just for them. It falls apart the moment someone in finance is asked to build the dashboard the entire finance team will use, with no sense of UX, no sense of the jobs to be done, none of the depth a trained designer builds up over years. We saw a version of this when anyone could throw up a Wix site with a bad template. We’re about to see it at scale, across entire organisations, with software instead of web pages.
So the designers who have something to worry about aren’t the ones leaning into any of this. They’re the ones treating their current workflow as the ultimate ceiling - the Figma file, the brand system, the process they already know how to run, even the ideas - and assuming that if AI touches that, there’s nothing left for us, the humans. There’s a lot left. Voice interfaces need design thinking. Agent-facing products need design thinking. Every team now building its own tools without a designer in the room needs design thinking it doesn’t have, whether it knows it or not.
I think about this a lot, partly because I’m on the wrong side of the gap myself - I can spot bad design, I just can’t build good design, and AI hasn’t changed that for me yet. I also understand the fear, because generative AI has already changed how we communicate and write and that’s largely how I’ve made a living over the last two decades.
But my tl;dr is basically… things are changing, and ‘good design’ has never been more needed.
If any of this interests you and you happen to be in the Newcastle next week, I’ll be talking about the impact of AI on design at the Design in Everything launch event on Monday night. Find out more at Design in Everything