How AI is Redefining the Design Process
What it means to design when anyone can build. Why functional prototypes are the new design standard in the age of AI
Design is being redefined at its core.
Generative AI has shifted the boundaries of what can be imagined, produced, and shipped — blurring roles, accelerating workflows, and collapsing the gap between idea and execution. What started as novelty: text-to-image, code completion, auto-translation, has quickly evolved into a fundamental reconfiguration of creative and technical work.
For designers, this isn’t just a tool upgrade. It’s a tectonic shift: in what the role demands, what it enables, and what it might soon make obsolete.
Do you remember your first encounter with generative AI?
For me, it was MidJourney. I still remember the moment I saw the quality of the art it could generate — I thought, OMG, I’m glad I dropped out of art school and didn’t become an artist.
Since then, everything’s been accelerating. AI isn’t replacing creative work one-to-one — it’s expanding the scope of what’s possible. We’re seeing sharper images, more localised content, an explosion of video, voice becoming a standard for both input and output.
In short: the pie got bigger. More output, higher quality, lower cost.
And what about the people behind the work?
The smartest artists, copywriters, developers, and translators aren’t competing with AI, they’re orchestrating it. They’ve become expert humans in the loop, or evolved into new roles entirely: brand strategist, creative technologist, content lead.
But not everyone is keeping pace.
So… what about UX?
I started out as a “webmaster” — designing and coding. That blend served me well early in my career. But as both fields matured, the industry split, and roles became more specialised.
Today, UX spans a vast range of skills: research, architecture, flows, microcopy, visuals, motion, design systems. That makes it hard to automate end-to-end. But individual parts are increasingly powered by AI. Or quietly handed off to others. PMs are assembling prototypes. Engineers are building polished UIs with design systems and prebuilt components.
From specialists to unicorns (again)
At the latest Stripe Sessions, John Collison shared how very small teams are now building remarkably solid products and businesses. Think of Cursor, that reached valuation of 300M with 60 people, Lovable and Bolt that are the fastest growing startups in the US and EU, while having teams of ~20 people. We're also seeing a wave of niche startups and solo ventures that wouldn’t have been viable in the old world — like a dancing network, or SaaS for yachting events.
This is possible because people are wearing more hats, enabled by AI, and spending less time on politics, meetings, and handoffs. Or as Claire Vo puts it: Super ICs — people who aren’t limited by conventional roles, and can adapt to what the situation demands. Sometimes they lead teams and shape strategies, and other times they build things hands-on.
The World Economic Forum predicts that by 2025, 50% of all employees will need significant reskilling. This is especially true for knowledge workers like us.
The shelf-life of professional skills is shrinking dramatically. The best way to stay in the game is to keep learning — and even harder — to unlearn what used to be true, but no longer is. We need to get comfortable with that.
“AI won’t replace people. But people who harness AI will replace those who don’t.”
— Ginni Rometty, Former CEO of IBM
Let’s talk UI design tools
If you’ve used DALL·E or MidJourney to generate UI, you know the results: well, interesting, but not particularly useful.
Figma’s 2024 Config AI launch? Underwhelming. Figma Make? Still TBD.
UX Pilot and Galileo? Promising, but haven’t taken the industry by storm.
But these tools did move the needle:
Lovable, Bolt, V0, Replit, Cursor, Windsurf.
These aren’t just design tools — they’re changing how we design.
Suddenly, building a functional prototype is faster than polishing a static Figma mockup.
Do we even need visual design before building?
We used to operate under the belief that coding was slow and developers were expensive. That mindset shaped an entire process: invest heavily in design, user testing, and stakeholder alignment — all before writing a single line of code.
But that logic no longer holds.
I used to say “Design is cheap, code is expensive.”
Now code is cheap, we can build, test, iterate faster and cheaper than before.
With a new wave of AI-powered tools, the cost of building has dropped dramatically. And more importantly, building isn’t just faster, it’s now accessible to anyone, even without writing code.
That shift raises a serious question:
Do we even need polished visuals before we start building?
My own design process changed the moment I realised I could build a functional app in Lovable faster than crafting a static flow in Figma.
Data and function are the experience
Interactivity, responsiveness, real-time logic, especially in this era of non-deterministic AI, are what shape meaningful user experience. Often, these matter more than visuals.
That’s why functional prototypes are such a game changer.
They push us beyond static screens — toward building behaviours.
They give us better testing, sharper insights, deeper collaboration with engineers.
Figma is the new napkin sketch
These days, I use Figma very differently. It’s become my sketchbook — chaotic, disorganised, full of fragments and ideas. I still love the infinite canvas, and for some things, the familiar act of “drawing” is simply faster.
But once the vision starts to take shape, I move it into an AI prototyping tool. That’s where I refine it, not just in how it looks, but in how it behaves.
Design isn’t about artefacts. It’s about experiences.
Design isn’t about creating artefacts — it’s about creating experiences. And today’s tools let us build actual experiences with just a few prompts and clicks. Real products, real data, real interactions.
Not using these tools isn’t just a missed opportunity. It’s almost negligent. We owe it to ourselves, our teams, and our users to get as close as we can to the real thing.
And if we can go one step further and build it ourselves? That’s a superpower.
How can we use AI prototyping in day-to-day design work?
Not every designer needs to build standalone apps. Most of us work on features inside complex systems, collaborating with engineers, PMs, and other designers. But that doesn’t mean AI prototyping isn’t relevant — far from it. Its potential use cases are broader than many realise.
I keep calling it AI prototyping for lack of a better term. But the name doesn’t quite fit. These tools often go beyond just prototyping. We don’t need to limit ourselveswith just mocking things up, we can build shippable products.
Still, I stick with AI prototyping because I don’t want to suggest that anyone can launch production-level software without any technical understanding. You do need to grasp how systems work. But AI lowers the bar dramatically, and makes it easier to learn by doing.
So how can we, as designers, actually use AI to build?
Here are four big use cases I keep coming back to — though I’m sure more will emerge.
1. Design
We might co-create and ideate new design concepts with it. These tools can generate high-fidelity, interactive, responsive UI — with or without initial Figma design. You can explore interface patterns and behaviours that go beyond what Figma can simulate.
2. Disposable Prototypes
We might test product ideas more effectively. Testing with functional apps will certainly give more reliable insights, than mockups stitched together. These are great for aligning with stakeholders or sparking brainstorming sessions in real-time.
3. Reusable Code
We might deliver production-grade components or frontend logic directly to developers and continue collaborating directly in code rather then letting details slip in translation, not mentioning how much longer it takes to explain what we want, rather than doing it ourselves. It enables working in a shared space where code can evolve across both design and development teams. I am incredibly optimistic about this scenario, as I believe this would allow designers accept less compromises, and developers recover time for solving engineering and architectural challenges, rather than fixing spacing and colors.
4. Production-Ready Apps
Last but not least, we might build and release new products from ground up. This opens so many doors to start new business or build software tailored to very unique needs.
Final Note
The choice of a tool, and how you prompt it, depends entirely on your use case.
If you’re approaching a clear, well-defined task with mockups or requirements in hand, treat it like a mini PRD. Be direct and structured in your prompts. Guide the tool toward precision.
But if you’re exploring new ideas or looking for unexpected directions, let the AI be your creative partner. Use open-ended, ambiguous prompts. Focus on the problem, not the solution. Let it hallucinate a little — that’s often where the interesting stuff emerges.
I’ll go deeper into tool comparisons and prompting strategies in future posts.
—> Comparison of Lovable, Bolt, Replit & V0
This article is based on a talk originally delivered at Tech Circus, Berlin, in May 2025.