Design and engineering shouldn't be a handoff

Most product teams are built around a wall. Designers decide how it should look and feel, then throw the work over to engineers, who decide how it actually behaves. It's been the default for decades. For AI products, it quietly breaks.

In AI, the design and the intelligence are one decision

With traditional software, you can mostly separate the two: the design describes the screens, and engineering makes them work. The behaviour is fairly predictable, so the handoff loses little.

AI removes that predictability. What the product does — how it responds, when it's confident, when it's wrong, how it recovers — isn't a fixed spec a designer can hand over. It's a moving, probabilistic thing that lives inside the engineering. So the most important experience decisions (how much to trust the model, where to keep a human, how to make an answer legible) can only be made by someone who understands both the interaction and the intelligence. Split that across a wall and the design describes a product that doesn't exist, while the engineering ships behaviour nobody designed.

The wall is where generic comes from

This is also, quietly, where sameness comes from. When design hands over a static picture and engineering has to make an unpredictable system fit it, the path of least resistance is the default component, the default flow, the safe pattern that "works." Nobody chose generic — the handoff produced it, because no single person owned the whole thing well enough to make it specific.

The teams that ship AI products that don't feel generic tend to have one trait in common: the person shaping the experience and the person building the intelligence are either the same person or sitting so close together that there's no wall to throw anything over.

What "fused" actually looks like

Fusing design and engineering doesn't mean everyone does everything. It means the experience and the implementation are decided together, continuously, by people who can hold both in their head:

• The AI's behaviour is designed, not discovered after the fact — because whoever's shaping the flow also understands what the model can and can't reliably do.
• Trust patterns, error states and human-in-the-loop moments are built in from the first sketch, not retrofitted when QA finds them.
• The thing that ships is the thing that was scoped, because there was no translation step to lose intent in.

The output isn't just prettier. It's more coherent — the product feels like it was made by one mind with a point of view, because effectively it was.

Why this is the moat, not "design"

It's tempting to say the answer is "be more design-led." It isn't, quite — design alone, handed over a wall, produces the same brittle result. The durable advantage is the fusion: judgment about experience and judgment about engineering, applied by the same team, to a medium where those two things are no longer separable.

That's harder to copy than a visual style and more valuable than a faster build, precisely because most teams are still organised around the wall. It's also why we run the studio the way we do — strategy, design taste and AI engineering in one team, so the product that ships is the product that was designed. That's the whole idea behind how we work — or tell us what you're building.