For years, we’ve described the design process as an orderly, almost comforting sequence: analysis, brainstorming, mapping, wireframes, prototypes, testing, iteration. A clean, progressive path where each phase set the stage for the next.

Then came AI. And suddenly, the challenge is no longer just managing to produce a solution. The challenge is figuring out whether that solution actually deserves to exist.

Did the old way work? Yes.
Does it still work? It depends.

But pretending that today, with AI baked into the creative and design process, we can just keep designing the exact same way is pretty naive. Or worse: convenient.

Because AI isn't simply an accelerator. It doesn't just exist to help us do what we did before, only faster. If we use it that way, we’re putting a Ferrari engine inside a Fiat Panda and then complaining that it doesn't change much.

This theme is already obvious in the broader discussion around product design. McKinsey, for instance, highlights how generative AI can significantly shorten design cycles, but stresses that it doesn’t eliminate the need for expertise, judgment, and discretion on the part of designers. Translated: speed is useful, but on its own, it doesn't know where it's going.

The real transformation is something else: today, we can start from the solution much earlier.Not from the definitive solution. Not from the perfect solution. But from something that exists—something you can see, critique, break, and improve.

And that changes everything.

From Linear Process to Reverse Process

The old model started from the problem and moved gradually toward the solution. First you understood, then you imagined, then you designed, then you tested.

Today we can do something different: quickly generate a first version, even a rough one, and use it as raw material for reasoning.

This doesn’t mean skipping analysis. It means stopping treating analysis like a locked room where you have to stay for weeks before producing a single thing.

An initial solution can become a better question.

A quickly generated prototype can bring to light constraints, inconsistencies, and opportunities that an abstract discussion would have never revealed.

In this sense, AI allows us to design in a way that is less linear and more exploratory. You start with a possible answer, then you go backward, understand what isn't working, redefine the problem, and refine the direction.

It’s a messier process. Less elegant on paper. But often, far more useful.

The Point Isn't to Create More Things. It’s to Choose Better What to Create.

Here comes the uncomfortable part.

If anyone can generate interfaces, copy, concepts, images, flows, and prototypes in just a few minutes, the differentiator is no longer just the ability to produce.

This isn't a theoretical scenario: in its 2025 AI Report, Figma describes a community of designers and developers increasingly involved in designing and building AI-powered products. Output production is becoming more accessible, faster, and more distributed.

The difference lies in the ability to choose which problem is worth tackling, which solution deserves to be explored, what to cut, and what not to build.

Because in a world where creating is easy, creating useless things is effortless.

And a wrong solution, beautifully executed, weighs far heavier than a good idea left in a drawer.

A bad idea that never hits the ground does little damage. A bad idea quickly turned into a product, a marketing campaign, or a user experience can generate confusion, costs, design debt, and mismatched expectations.

AI lowers the barrier to execution, but precisely because of that, it raises the stakes for decision-making responsibility.

AI Shouldn't Just Be the Muscle. It Needs to Be in the Brain of the Process.

Using AI only for execution is limiting.

“Write this text for me.”

“Generate this screen for me.”

“Give me ten ideas.”

“Turn this wireframe into UI.”

All useful. But it’s not enough.

If AI remains just an operational tool, it becomes a very fast assistant producing outputs based on decisions already made. The real leap happens when we involve it earlier: in defining alternatives, challenging assumptions, evaluating trade-offs, and simulating scenarios.

Not to delegate judgment to it, but to elevate the quality of our own.

AI can help us spot blind spots, stress-test a choice, generate counter-arguments, compare different directions, and highlight inconsistencies between business goals, user needs, and technical constraints.

In short: it shouldn't just help us build faster. It needs to help us decide better what to build.

Create Right Away, Yes. But Create Well.

There is a dangerous misunderstanding: because we can produce something immediately today, people think doing so is always a good idea.

No, it isn't.

Starting from the solution doesn't mean throwing anything out there and hoping that iteration will work miracles.

An initial solution must be concrete enough to be discussed, but also flexible enough to be modified. It must support the refinement process, not block it.

The Nielsen Norman Group also urges us to look at AI-assisted prototyping through this lens: tools that are useful for rapid exploration, but dangerous if mistaken for finished products or shortcuts capable of replacing design thinking.

If I create something that is too rigid, too polished, or too in love with itself, I'm just fast-tracking the moment when it becomes difficult to change.

The first output shouldn't look final. It must be designed to be questioned.

It should help the team say:

That is the point: AI shouldn't just lead us to succeed faster. It should lead us to fail faster.

And above all, to fail when it’s still cheap.

The New Value of the Designer

In this scenario, the designer doesn't lose centrality. However, the job changes.

Less a patient executor of a predefined process. More a director of decisions, constraints, possibilities, and consequences.

Less obsessed with checking off all the "right" steps. More capable of understanding which steps are actually needed, in what order, and to what depth.

The design process doesn't disappear. It becomes more modular, faster, and more critical.

Sometimes it will start with a map, sometimes with a prototype.

Sometimes with a provocation generated in half an hour, and other times with a wrong solution that finally makes the right problem obvious.

The differentiator will rely less and less on the ritual of the process and more and more on the quality of the decisions—because AI can help us create almost anything, but it cannot yet save us from creating the wrong thing with great efficiency.

And this, perhaps, is the most important new skill: no longer designing slower to feel safer, but designing faster without becoming superficial.