Engineers go first

Matt Steenburg, Fantasy’s Head of Design, on how the design process is changing with AI: engineers set the foundation first, strategy and high-craft design stay human, and clients click a real, working v0.1 in the first days of an engagement instead of approving static work for months.

Two years ago, launching a product, even for a startup, usually meant six to eight months before anything foundational was live. Clients rarely pushed back on that timeline, because the process behind it worked. It still works. It’s slow, and the appetite for slow is gone.

For a long time the responsible way to build was to learn everything up front. Do the rigorous testing, make sure it’s the right thing, then go build it. I’ve spent my career leading design teams through that sequence, and I’ve watched us do all of it faithfully and still be wrong. What AI changes is that you can put a real thing into the world and learn from it while it’s live. You adapt in real time instead of betting six months on your research being right.

Acting on that changes how a team has to work, starting with who goes first.

Engineers go first. That’s a strange sentence for a head of design to write, but I mean each word. Today our teams are setting the foundation and standing up the working environment before design has finished a single screen, so there’s a real place to build from day one and the beginnings of a design system out of the gate.

Strategy stays human-led, and I don’t think that should ever change. We still get in a room and argue our way to a shared answer on what we’re building and why. You could let AI do that part too, but then are you actually in agreement that it’s the right thing? Sitting down and being human together is still where the work starts, and that part doesn’t need to be fast. What’s changed is what happens the moment it’s done: we agree on the strategy, then on the technology, and every discipline starts together on day one.

Designers work in the same space the engineers will ultimately pick up. One team working end to end, with decisions landing in one place, instead of five teams doing their own thing in their own tools and reconciling it all at the end.

Early on, the screens are deliberately plain. Structure without styling, flows without a visual language, so the team can argue about whether the product works before anyone falls in love with how it looks. The polish comes later, and it comes on purpose.

The design system is the clearest example of what all this changes. For most of my career, a design system was something you assembled after the product was designed, then maintained by hand forever. Now the system grows while the product grows. Every decision gets absorbed as we go, and by the time the visual direction locks, the system is already there waiting for it.

The strategy and the craft. We hold a very high bar for visual craft at Fantasy, and AI can’t hit it, so high-craft visual design stays manual on purpose. Parts of the process accelerate, mood boarding and visual exploration especially, but in the end a person sits down and designs the screens. Once a direction locks, the system is what should let that craft spread through the whole product, fast.

In one of our internal demos, a working website went from prompt to clickable in about ten minutes. Ten minutes gets you a demo, and a demo is a long way from a product. That kind of raw speed is a parlor trick until you connect it to a real foundation and real judgment. Then it’s a process.

Anyone who has tried to prompt an entire app into existence knows what happens without taste and quality control in the loop: you iterate on something that’s constantly almost falling apart, and you can end up with a monstrosity. Human judgment is what makes the speed worth having. That’s true for design, and just as true on the engineering side, where senior people still review and sense-check the work, especially when the product carries real constraints around compliance or security.

Clients stop reviewing artifacts and start using the product. The old sequence asked a client to approve an art direction, then a UX, then wait while everything got assembled and built. Now, in the first days of an engagement, a client is clicking through something that looks and behaves like a real product. Some paths dead-end. It’s a v0.1, and it’s real, and it changes at the speed we make decisions together.

Prototypes always promised this. They couldn’t deliver it, because building them was slow and manual: you’d invest weeks in something everyone knew would be thrown away. When the prototype lives in the same environment the engineers build in, the thing a client reacts to in week two is the same thing that keeps growing toward launch. The limiting factor stops being production and becomes the pace of decisions, ours and the client’s together.

And when we hand the work back, the client should inherit the way of working along with it: an accelerated process their own team can pick up and keep iterating. That handoff is the part I most want to get right.

Everything above is the model as we run it. It’s fair to ask how much of that is proven, and the honest answer is that we’ve been running it in pieces so far: on internal builds and on parts of client engagements. Right now we’re proving out the full system, and the hardest part is exactly where you’d expect, the climb from a solid working foundation to the level of craft we’d put the Fantasy name on. While the gap is real, we’re working on closing it, and I’d rather say so than pretend the whole thing arrived finished.

It’s also best suited to a singular platform right now: one app, one website. Stitching together an ecosystem of hundreds of products would mean rethinking the approach, and we haven’t taken it there. If your world is that ecosystem, the honest answer from anyone selling this kind of process should include that limit. Better to say it in the first meeting than have you discover it in month three.

There’s a learning curve for people too. Design and engineering stay distinct disciplines, and they need to work closer together than they ever have, close enough to be jamming on the same thing at the same time. That’s a real change in habits, and I suspect it’s the most interesting thing happening to either discipline right now. That one deserves its own piece.

I don’t see a world where this direction reverses. Clients are already asking for it: they want the accelerated process, they want to do more with fewer people, and they want to see a real thing early. Our job is to give them the speed without giving up the judgment. The part I keep circling back to is the prototype itself, and what it becomes when it stops being disposable. That’s where I’ll pick this up next.