I spent years managing design teams without actually making anything. Then AI arrived, and I started building again — using Claude to turn ideas directly into working prototypes. No briefs, no handoffs, no waiting. Just thinking something and watching it appear on screen.

It felt great. Until it didn’t.

When I finished, I stared at what I’d built and realized I had no idea why I’d made it. The design was solid. The prototype worked. It could probably ship. But somewhere in the excitement of making, I’d forgotten to ask the only question that actually matters:

If this succeeds, what changes?

Design process exists for real reasons. It creates alignment, reduces miscommunication, and gives teams a shared language for moving forward. When Jenny Wen, Anthropic’s Head of Design, declared that “the design process is dead,” I actually pushed back. My instinct was to defend it — process isn’t just bureaucracy, it’s the infrastructure that lets design work inside organizations that don’t naturally speak design.

But I’ve been sitting with something since then.

Process has a quiet side effect that nobody talks about: it creates the illusion that direction will emerge if you just follow the steps. Keep moving through the phases, and the right answer will reveal itself at the end.

It won’t. Direction doesn’t emerge from process. It has to be defined before process begins.

For most of design history, this wasn’t a crisis. Execution was expensive. Every step cost time and money, so the discipline of working through a process protected teams from costly mistakes. The process was load-bearing.

AI changed the load-bearing equation.

Now anyone can generate ten concepts, ten interfaces, ten prototypes before lunch. The bottleneck was never making things — it was always knowing which things were worth making. AI just made that gap impossible to ignore.

In 2017, Pentagram’s Natasha Jen stood on the 99U stage and called Design Thinking bullshit. She wasn’t wrong, but her argument was aimed at the wrong target. The real problem wasn’t that non-designers were using design process badly. It was that designers themselves had stopped questioning whether the process was pointing them toward the right problems.

Seven years later, two things have happened that make her warning feel prescient in ways she didn’t anticipate.

The first is what designer Felix Lee calls the code designer — a shift where designers don’t just describe products, they generate them. When design and engineering collapse into the same motion, production speed increases by orders of magnitude. The gap between idea and artifact is nearly gone.

The second is what happens to thinking when execution becomes effortless. The old cost of defaulting to process was wasted time. The new cost is an entire team moving very fast in a direction nobody stopped to question. Cognitive inertia has always been expensive. At AI speed, it’s catastrophic.

Here’s what I’ve started noticing in organizations.

A project kicks off. Timelines get confirmed, resources get allocated, deliverables get defined. The machine starts moving. From the outside, it looks like progress.

But ask one question — if this project succeeds, what will users actually do differently? — and the room goes quiet.

Not “improved experience.” Not “better interface.” A specific, observable change in behavior. Most teams have never discussed it, because process gave them permission not to. If we follow the steps correctly, the right outcome will be waiting at the end. That’s the deal process implicitly offers. And for a long time, it was good enough.

AI broke the deal. When you can execute ten times faster, the cost of moving in the wrong direction is ten times higher. Output goes up. Clarity doesn’t follow automatically. Someone still has to own the result, and that’s not something you can delegate to a model.

At Stanford GSB, I heard innovation described in terms of friction — and that friction is always defined by outcomes. Without a clear outcome, everyone on a team interprets success differently. Friction multiplies. You ship things that don’t add up to anything.

This is the one place AI genuinely can’t help. It can generate. It can execute. It can accelerate. But it cannot tell you where you’re trying to go. That’s still entirely a human problem.

Which means the most important design skill in an AI-accelerated world isn’t knowing the process. It’s being able to walk into a room before a project starts and answer, clearly and specifically, what success looks like.

Before your next project kicks off, try replacing the first agenda item. Instead of “how do we structure this,” ask: what does the world look like if we get this right?

Forget the process. Start with the outcome.