Design Is Back. The Job Description Has Changed.

Last week, we hosted an evening with Figma in Milan called Design is Back. A room full of design leaders, two talks from Figma, a short one from me, and an aperitivo that lasted considerably longer than the presentations.

The title was deliberately provocative. Design never really went anywhere. But for a while, it felt as though the conversation had become more about the tools than the people using them. Then AI arrived, made those tools accessible to almost everyone, and something interesting happened.

The easier it became to make things, the more obvious it became that making things was never the hardest part.

That’s what the evening kept coming back to, from three different directions. And I wanted to write some of it down before it fades, because Riccardo and Ron from Figma managed to put words to a few things I’ve been wrestling with at our studio for the past year.

We’re producing more. But are we designing better?

I opened my talk with a frustration I’ve been carrying around for a while.

AI has given almost everyone the ability to produce something that looks designed. A product manager can describe an idea and get a working prototype. A marketing team can build a landing page before lunch. Someone who has never opened a design tool can produce an interface that looks perfectly respectable.

I genuinely think that’s wonderful. We’ve spent the past two years pushing AI adoption across our studio, and I have no desire to go back to the days when every small change required a specialist and three rounds of handovers.

But there’s a difference between making design accessible and making good design inevitable.

I’ve started calling the result Fast Food Experiences.

They’re consistent. They’re familiar. They arrive quickly. And they all taste vaguely the same.

You know the ones. Beautifully rendered interfaces that look like every other interface generated that month. Impressive at first glance, but strangely anonymous. Nothing is technically wrong. The spacing is fine, the hierarchy makes sense, the components are polished.

And yet, you couldn’t tell which brand they belong to.

The problem isn’t a lack of production quality. It’s a lack of authorship. The result looks finished, but you struggle to identify the decisions that made it distinctive.

For a while now, I’ve been telling my team that our job is moving toward orchestration, not decoration.

That evening, I discovered two rather compelling ways of explaining why.

Speed is no longer the competitive advantage

Riccardo, Developer Advocate at Figma, opened with three words that sound obvious until you consider how rarely they coexist.

Speed. Direction. Differentiation.

Without speed, you’re too late. Without direction, you’re just producing noise. Without differentiation, nobody remembers what you made.

AI has made the first one dramatically easier. We can describe an idea and watch something appear on screen. And because that moment still feels like magic, it’s where most of the attention goes.

But the magic is becoming ordinary.

When anyone can build something quickly, speed stops being the thing that separates you from everyone else. The difficult questions move elsewhere.

What should we actually build? How do we know it’s the right thing? How does it fit with everything we’ve already created? And perhaps most importantly, why should anyone choose our experience over another?

Riccardo made another point that resonated with me. Teams often organize themselves around separate activities, but customers don’t experience separate activities. They experience one product.

You might have a brilliant design team, a fast development team, and an AI agent producing prototypes overnight. But if each works with a slightly different understanding of the problem, you’re not moving faster. You’re just creating more expensive misunderstandings.

Every handover becomes an opportunity to lose context.

The reasoning behind a decision disappears between the brief, the design file, the prototype, and the production code. Someone downstream has to reconstruct it, or worse, makes a different decision without realizing it.

Anyone who has inherited a Figma file containing forty frames and absolutely no explanation knows the feeling.

And I liked a smaller observation Riccardo made about how people work.

Sometimes you want to describe what you need. Sometimes you need to interact with a working prototype to understand whether an idea makes sense.

And sometimes you just want to move a bloody rectangle two pixels to the left.

Anyone who has written five prompts to fix a two-pixel alignment knows exactly what he meant.

The future of design tools isn’t about forcing everything through a prompt. It’s about giving people different ways to think, explore, and make decisions, without losing the context along the way.

Figma is building around that problem

This is where the product announcements started making more sense to me.

Riccardo walked us through four updates from Config India in Bengaluru, which had taken place a couple of weeks earlier. Individually, they’re interesting new capabilities. Together, they suggest something more ambitious.

Code Layers, now in open beta, bring working prototypes directly onto the Figma canvas.

Instead of treating code as something that happens after design, you can explore functional interactions alongside visual concepts. Compare approaches. Test ideas. Discuss a working experience in the same environment where you’re making decisions about it.

Code becomes another material to think with, rather than simply the final output of the process.

Then there’s a dedicated space for design systems, shown as a preview.

This one is particularly relevant to the work we’re doing at Song. The idea is to bring component libraries, guidelines, and code into a connected environment where teams can understand not just what exists, but how everything relates.

And here’s the part I find genuinely important: it becomes a reference not only for people, but for AI agents.

Instead of generating yet another plausible-looking component from scratch, an agent can work from the actual decisions a brand has already made.

The difference is enormous. You’re no longer asking AI to invent what your product should look like. You’re giving it the knowledge required to extend something that already has an identity.

Figma Specs, also in open beta, address the other half of the problem.

A design system tells you how something should be built. A spec carries the context explaining why you’re building it in the first place.

The brief, the requirements, the research, the decisions and constraints. The information that tends to disappear as a project moves from one person or tool to another.

And finally, Payload, the open-source headless CMS integrated into Figma Make, brings real content management into the picture. Because a working experience needs more than a convincing first prototype. It needs content that can evolve without rebuilding everything around it.

Figma calls the broader idea the intelligent canvas.

I’m usually a little suspicious of platform terminology, but I like what this describes.

A shared space where ideas, designs, code, context, and conversations remain connected. Where people can contribute their expertise, agents can handle parts of the execution, and the reasoning doesn’t disappear every time the work changes hands.

It sounds less like a new tool and more like a better way of working.

And it leads directly to the question that shaped the second half of the evening.

If AI can take away so much of the execution, what should we actually keep doing ourselves?

Give away the labor. Keep the thinking.

Ron Baranov, Creative Workflow Advocate for Figma Weave, came at this from a completely different angle.

Ron spent ten years in video production and the last two creating with AI. His presentation was wonderfully visual, with almost no words on the slides, and he delivered some of my favorite lines of the evening.

He started with a breakdown he credits to Gary Vee.

Roughly 15% of creative work is deciding what to make. Another 80% is making it. And the final 5% is finishing it.

Obviously, those aren’t scientific measurements. But anyone who has worked in a creative studio recognizes the proportions.

For years, that 80% consumed most of our time.

The production. The revisions. The rework. Exporting the same campaign in forty different formats at midnight. Making another version because someone changed their mind about the background.

AI is compressing that part of the process at an astonishing rate.

And in theory, that’s great news. More time for ideas. More room for experimentation. More energy available for the decisions that actually matter.

Except there’s a catch.

Not all friction is wasted effort.

Some of the best creative work comes from the struggle. The terrible first version that eventually leads somewhere interesting. The accidental result you almost deleted. The uncomfortable conversation where someone challenges a direction everyone else had accepted.

Those moments feel inefficient. But they’re often where the real thinking happens.

And they’re also the first things we’re tempted to automate.

Ron summarized the distinction beautifully:

Give away the labor. Keep the thinking.

Because if you give away the thinking as well, all you’ve accomplished is getting faster at producing forgettable work.

He illustrated it with the old green-screen production process. Tracking, pre-visualization, 3D rendering, compositing, edge refinement. An enormous amount of technical effort just to arrive at an image that feels convincing.

AI can now eliminate much of that machinery.

Should we miss it?

Not necessarily.

Ron shared a thought from Itay Schiff that stayed with me: we shouldn’t preserve friction simply because we once called it craft.

That’s a surprisingly uncomfortable idea for creative people, because so much of our professional identity has historically been tied to mastering difficult tools and processes.

But difficulty isn’t inherently valuable.

The important distinction is between friction that prevents you from creating and friction that helps you create something worth making.

For most of our careers, the technology decided where that friction lived.

Now, increasingly, we get to decide.

And that changes the role of the designer.

Less operating tools, more directing systems. Less repetitive execution, more judgment. Less obsessing over how to produce an output, more attention to whether that output deserves to exist.

Ron demonstrated this with a creative brief that generated a mountain of alternatives.

Only four survived.

The others weren’t necessarily bad. Some were technically excellent.

They just weren’t right for the brand.

And that is the part I keep coming back to.

Generating a hundred options is becoming trivial. Knowing which four belong together, which one expresses the idea, and which ninety-six should never see the light of day is a different kind of skill.

It’s taste. But taste informed by context, experience, and a clear understanding of what you’re trying to communicate.

From creative outputs to creative systems

This is also where Figma Weave enters the conversation.

Weave is a node-based creative canvas that brings together major AI models for image, video, 3D, and audio generation. It includes compositing and timeline capabilities, allowing teams to move from exploration to production in the same environment.

That’s useful, of course.

But what interested me most wasn’t the ability to generate more content.

It was the possibility of turning a creative direction into a repeatable system.

Think about how creative teams traditionally work. A creative director establishes a direction. A few experienced people understand it deeply. The rest of the team tries to reproduce it across channels, formats, and markets.

Much of that knowledge remains implicit. It lives in people’s heads, in references shared during meetings, or in feedback scattered across dozens of files.

With a node-based workflow, some of that reasoning becomes visible.

A decision can become a node. An approved approach becomes a workflow. The workflow can be packaged into a tool that other people use without having to reconstruct the entire process.

You design the system once, and the team can use it throughout a campaign.

Ron shared examples of what that can mean commercially: production timelines dropping from six to eight weeks to three, localization and social production costs falling by around 80%, and photoshoots that would normally cost $80,000–90,000 being replaced by a day of AI production.

Those are compelling numbers.

But I think the bigger change is cultural.

Creative quality has traditionally depended heavily on the people doing the work at any given moment. We’re beginning to see how parts of that knowledge, those decisions, and even elements of creative judgment can be captured in systems that others can build upon.

Not to eliminate creativity.

To make sure every new piece of work doesn’t have to start from zero.

Ron closed with a fashion analogy. In Milan, that’s almost unfairly effective.

Somewhere in this city, a designer is probably cutting a look they spent months perfecting. It’s beautiful. It works. But it doesn’t belong in the collection.

Coco Chanel is often credited with the advice to remove one accessory before leaving the house.

AI can give you a thousand looks.

It still won’t tell you which one should walk.

What I’m taking back to the studio

I went into the evening expecting to hear about new tools. I came out thinking about how we should organize our work.

There are three things I want to carry forward.

First, AI adoption is no longer the hard part.

Two years ago, getting a hundred designers to experiment with AI felt like a serious challenge.

Today, we have an internal AI leadership think tank, and every two weeks we run a session called Play, where people share what they’ve built, what they’ve tried, and what they’ve learned.

People are experimenting. They’re building. They’re genuinely excited.

The problem has shifted.

We now have lots of isolated experiments, each with its own prompts, assumptions, and context. And alongside the genuinely exciting work, there’s a growing amount of workslop: content and prototypes that look finished but don’t really move anything forward.

We don’t need more generation for its own sake.

We need better connections between experiments, shared design principles, reusable knowledge, and a clearer understanding of what good looks like.

That’s why I’m increasingly interested in AI-ready design systems, specifications that travel with the work, and what we’ve been calling a Brand Brain: a shared intelligence layer that makes brand knowledge, decisions, and context accessible to both people and agents.

Second, friction is becoming a creative choice.

Ron’s framing gave me language for something we’ve been doing instinctively.

We can automate the midnight resizing without losing anything meaningful.

We can generate forty variations in minutes and still spend an hour arguing about which four are right.

We can accelerate production while deliberately slowing down at the moments where judgment matters.

The goal isn’t to remove every obstacle between an idea and an output.

It’s to recognize which obstacles are useless and which ones are actually helping us think.

Third, the most valuable thing we can scale is judgment.

This is where both talks connected for me.

Riccardo was talking about preserving decisions and context across product development. Ron was talking about embedding creative direction into reusable workflows.

Different domains. Same underlying principle.

The value isn’t just in producing something good once.

It’s in understanding why it’s good, capturing that understanding, and allowing other people and agents to build upon it without losing what made it distinctive.

A design system that contains actual decisions rather than just components.

A creative workflow where the brand’s taste is embedded in the process.

A shared memory that outlives the project and the people who worked on it.

That’s what I mean by orchestration.

Decoration is what the model can do on its own.

And since my dog Alfie has somehow become the unofficial mascot of my talks this autumn, I’ll offer a less sophisticated analogy.

Alfie doesn’t need me to walk. He’s perfectly capable of doing that himself.

He needs me to decide where we’re going.

Give him the leash and we’ll visit the same three lampposts for the rest of our lives.

The parts worth keeping

There’s an irony in all of this.

For years, we tried to make design more efficient by removing unnecessary steps. Now that AI is accelerating so much of the work, we’re realizing that some of those steps were where our best thinking happened.

The challenge isn’t to preserve our old processes out of nostalgia.

It’s to be much more deliberate about where we put our attention.

Ron’s final slide asked a question I think every creative leader should spend some time with:

Now that making is cheap, what are the parts you want to keep?

Not necessarily the parts you’re best at.

Not the parts that took years to master simply because the tools were difficult.

And certainly not the parts that used to justify our day rates.

The parts where the work becomes yours. Where taste, experience, context, and judgment turn something technically correct into something worth remembering.

Find those parts. Protect them. Build your processes around them.

And let the machines take care of the rest.

Design is back. It just came back with a different job description.

If you were in the room in Milan and came away with a different list, I’d genuinely love to hear it. Reply here or find me on LinkedIn.