What AI Can't Replace in Design

We've already written about how AI fits into our process at Hatch, and we meant it. We use it daily, for research, for moodboarding, for content creation and asset production at scale.

But using a tool well also means being honest about its limits. So here's the other half of that conversation: what AI genuinely can't do, and why that's exactly where the value of a designer lives.

Taste isn't a dataset

AI is exceptional at producing options. What it can't do is know which one is right.

That distinction matters more than it sounds. To be clear, that's not how we work, a logo isn't something we'd ever generate and pick from a lineup. But the principle holds across the process regardless of what's being produced. Whether it's a layout, a colour palette or a piece of creative direction, having more options on the table isn't the hard part. Knowing which one actually serves the brand, which one will age well, which one says something true about the business behind it, requires a kind of judgment that isn't generated, it's developed. Through years of looking at work, making decisions, getting things wrong, and building an instinct for what's right that's almost impossible to fully explain even to yourself.

That's taste. And taste is the thing AI has no real access to, because it isn't pattern recognition. It's conviction.

Why a typeface "feels right" isn't a technical answer

Ask a designer why a particular typeface works for a brand and you'll rarely get a purely technical answer. It's not just about x-height or letter spacing. It's about what the shape of the letters communicates, how it sits against the brand's tone, whether it feels confident or quiet, modern or timeless, and whether that's actually what the business needs to say right now.

AI can tell you what's popular, what's trending, what other brands in a sector are using. It can't tell you why one choice feels true to a specific business and another doesn't. That judgment call, made by someone who has sat with the brand, understood its ambitions and absorbed its context, is not something a model can replicate. It can only ever approximate.

Understanding the problem behind the brief

A client rarely arrives with the actual problem clearly stated. Often what they ask for and what they need are two different things, and figuring out the gap between them is a huge part of what good design work actually involves.

That requires asking the right questions, reading between the lines, and understanding a business well enough to know what's really going on. It's relational, not computational. AI can help process information once that understanding exists, but it can't build the understanding itself. That comes from conversation, context and genuine curiosity about a client's business, the things a designer brings to a project long before any visual work begins.

The risk of skipping the thinking

The real danger with AI in design isn't that it replaces designers outright. It's more subtle than that. It's the temptation to let it skip the thinking that good design depends on.

Work produced this way often looks plausible. It can look polished, on-trend, technically competent. What it tends to lack is point of view. It says nothing in particular, because nothing in particular was decided. That's the gap between something that looks like design and something that actually is design, and it's a gap that's easy to miss until you've seen enough of both to tell the difference.

Where the value still sits

None of this is a case against using AI well. It's a case for being clear about where the real value in design work comes from.

It comes from judgment. From taste built over years of paying attention. From understanding a client's business deeply enough to make decisions on their behalf with confidence. From the accumulated instinct that tells a designer when something is nearly right but not quite, and how to close that gap.

Tools will keep improving. That instinct won't be replaced by them. It'll just become more valuable, because it's increasingly rare, and increasingly the thing that separates work with a point of view from work that simply exists.That's what we bring to every project at Hatch, and it's not something we're willing to outsource.