My background isn’t digital. Before I spent 25 years solving digital problems, I was a real-world architect. And one of the things that has never stopped amazing me (across both industries, across three decades) is how consistently people try to skip the design phase.
In construction, it happens on the residential scale. Someone decides they know what they want, they find someone who’ll draw up plans without asking too many questions, and they build. The project becomes a mess. It’s been happening since before I qualified, and it’s still happening now.
In digital, it’s worse. Because at least in construction, there are regulations. You submit your drawings to the local authority. They get approved, or they don’t. There’s a forcing function that makes the design phase impossible to skip entirely.
Digital has no equivalent. Nobody is submitting their digital design drawings to the local council before the build starts.
Although, and I say this as someone who grew up in South Africa, given the state of our local councils, perhaps that’s for the best.
What I actually mean by design
When I talk about design, I need to be clear about what I mean. I don’t mean visual design. I don’t mean making it look nice or choosing the right colours.
I mean holistic digital product design. How the thing actually works. How easy it is to use. How well it’s technically architected. How thoroughly it meets the real needs of the people who’ll use it, not the people who imagined what those needs might be.
Just like a building isn’t just what it looks like, a digital product isn’t just what it looks like. The design is everything underneath the surface. And when that gets skipped (or rushed, or substituted with something that looks like design but isn’t) the project is already in trouble before the first line of code gets written.
What passes for design in most projects
Here’s what passes for design in most corporate digital projects:
Business requirements. Business specification. Technical evaluation. Quote. Build.
User requirements do get brought in more than they used to. I want to be fair about that. But how they get brought in matters enormously.
The most dangerous version I see regularly is customer journey mapping without real customers. Businesses will sit in a room (smart people, well-intentioned people) and map out what they believe the customer experience should be. It feels rigorous. It produces documents. It has sticky notes on walls and workshops and sign-off meetings.
But if real customers aren’t in the room, what you’ve produced isn’t user research. It’s internal assumptions dressed up as user research. And that’s actually more dangerous than doing no user research at all: because you think you’ve done it. You’ve ticked the box. You’ve given yourself permission to move to build.
How AI made building before designing worse
Now here’s where it gets interesting. And a little uncomfortable.
I’ve been using AI as part of my workflow for a few years. And what I’ve noticed, in myself, not just in clients, is that AI hasn’t solved the building-before-designing problem. It’s made it worse.
AI makes the build so fast and so cheap that it collapses the distance between idea and execution. Prototyping has become almost indistinguishable from building. And so the temptation to skip the thorough design phase and just start making something is stronger than it has ever been.
I catch myself doing it. Me, with 25 years of hard lessons cooked into my brain, with an architectural training that instilled design discipline before digital was even a career. And I still catch myself thinking: just start, you can fix it as you go.
Sometimes I’ll start a project from scratch when I realise I’ve done this. In my world, the rework cost is relatively low. But the codebase becomes a mess. And I’m working alone.
In a real business (with a real team, real infrastructure, real money, and a timeline that has commitments attached to it) that same AI-enabled shortcut creates a mess that is genuinely very hard to escape. The speed that AI gives you on the way in becomes the trap that holds you on the way out.
AI makes you feel like you’ve done the design work
But here’s the thing that concerns me most.
It’s not just that AI makes it easy to build before designing. It’s that AI makes you feel like you’ve done the design work when you haven’t.
Chatting to an LLM is a remarkably validating experience. It builds on what you give it. It encourages. It elaborates. It makes you feel like a superhuman: like you’ve covered every angle, considered every implication, thought it all through. Because AI talks to you as though you have done the work.
The danger isn’t that AI makes people lazy. It’s that AI makes people feel thorough when they aren’t. And that’s a subtler and more insidious problem: because you don’t know what you’ve missed. The unknown unknowns feel known.
Businesses are right now commissioning real digital projects using AI, believing that the speed and fluency of the process means the design thinking has been done properly. In many cases, it hasn’t. The foundation is incomplete. And the build is sitting on top of it.
The design discipline has to come from people
The discipline that the regulations provided in construction (the forcing function that made you stop and design before you built) has to come from somewhere in digital. It doesn’t come from the tools. It doesn’t come from the AI. It has to come from the people involved having the judgment to slow down before they speed up.
That judgment is built from experience. From having seen what happens when you skip it. From having rebuilt things from scratch enough times to know it’s always more expensive than doing it properly the first time.
Twenty-five years of that experience tells me: the design phase isn’t a cost. It’s insurance.
Garth Shoebridge is a digital problem-solver with 25 years of experience, and an architectural background that makes him take design seriously. If you’re about to start a digital project and you’re not sure whether you’ve done enough thinking before you build, start with a conversation.