Modern Skill Stack — Real stories on how product professionals are keeping up with a constantly changing skills landscape. Hosted by Gene Kamenez, CEO of Uxcel.
Listen on YouTube.
I sat down with Florian Boelter, a Staff Product Designer at Juro, a legal-tech platform serving in-house legal teams, where he has spent 3 years leading AI adoption across the design organization. Outside his day job, he runs Open Doors, a community that connects junior designers with job opportunities, portfolio feedback, and hiring insights every week. His sell-out workshops on AI in UX and vibe coding for designers on Uxcel consistently rate above 4.5.
What makes Florian’s perspective different is that he sits at three intersections most designers don’t: building AI features inside a real product, experimenting with AI tools to build his own side projects, and advising hundreds of junior designers on how to break into the industry right now.
So when I asked him what skill he’s personally focused on learning today, his answer wasn’t what I expected.
This episode is brought to you by Uxcel.
Uxcel is a learning platform used by 500,000+ product professionals and 200+ organizations to build cross-functional skills across UX design, product management, and AI. Learn through short, gamified courses that fit into a working day, earn certifications, and build a portfolio with real-world project briefs.
Start learning at uxcel.com
What you’ll learn
Why a staff product designer is doubling down on fundamental UX skills, not AI tools right now
The three-layer AI skill stack for designers and PMs in 2026: from table stakes (LLMs) to interesting (vibe coding tools) to advanced (Cursor, Claude Code), and which tier you actually need
How AI in products is shifting from sloppy to truly embedded features, and what that means for how you design them
What happened when Florian pitched AI-assisted design system work at Juro in 2023, and why the engineers moving to Cursor six months later changed everything
The exact portfolio gap separating junior designers landing roles at cool companies from those who keep getting ghosted
Why the product triad of design, PM, and engineering is getting closer, and the cross-functional skill that’s driving it
Timestamps
0:00 Introduction
5:00 The importance of design fundamentals
9:49 Role of AI in design and user experience
13:34 Building AI features: necessity or nice-to-have
19:05 Misconceptions about AI in design
22:28 Do designers need to learn to code?
25:06 The role of systems thinking in design
27:20 Empowering the product triad
31:21 AI tools and skills bare minimum
34:21 Building internal tools with AI
45:22 Leading AI adoption in organizations
51:27 Which designers are thriving vs. which are struggling right now
56:38 The importance of foundational skills
1:02:19 Advice for staying relevant in design
1:04:22 Outro
My biggest takeaways
The differentiator with AI isn’t an AI skill, it’s what you can imagine
When I asked Florian what skill he’s personally focused on right now, I expected him to name a tool or a model. Instead, he said typography, color, and composition. Not because he lacks confidence with AI (after all, he’s been building with it since the early days) but because he’s identified where the real gap will open up.
In Florian’s words:
“If I can prompt AI to do something, it’s not a matter of the output itself that’s produced. It’s more a matter of what I can actually imagine and how I can describe what needs to go in. So the output eventually becomes something nice and novel. That’s why I’m trying to mostly focus on scaling up in craft; because I otherwise feel quite capable in the AI department already.”
This is the clearest framing I’ve heard of why craft matters more, not less, in an AI-assisted workflow. You’re not competing with the tool. You’re competing with other people who are also using the tool, and the gap between them is the quality of their vision and the vocabulary they use to direct it. When everyone can generate, the one who can imagine well and describe precisely wins.
The action step here is counterintuitive: if you feel behind on AI skills, stop and check whether a foundation problem is underneath that. Florian argues that fixing the foundation pays a higher return than chasing the next tool.
LLM fluency is table stakes. Vibe-coding tools are where it gets interesting.
Florian draws a clean line between what’s mandatory and what’s still emerging. Using ChatGPT, Claude, or Gemini to accelerate day-to-day work (e.g. summarizing user feedback, drafting copy, getting quick context on something) that’s no longer optional for designers or PMs, or frankly for anyone in a knowledge-work role.
In Florian’s words:
“A basic command of ChatGPT, Claude, or Gemini — I feel like that’s definitely a mandatory skill that even goes beyond design into non-tech jobs at this point. But the more interesting place is where AI is embedded in areas and functions where you can’t even see it’s AI. It’s just helping you get the job done. That’s where the complexity is, and that’s where it’s actually getting interesting.”
The practical split Florian recommends: once LLMs are part of your daily workflow, the next layer is picking one higher-level tool such as Lovable, Figma Make, or similar, and spending a focused weekend with it. Build something you’re actually concerned with at work. The goal isn’t to launch anything. It’s to understand what’s possible before you need to explain it to anyone else.
Beyond that, the tools diverge by role. If you work heavily on design systems, tools like Claude Code and Cursor are already mature enough to be worth the investment. If you’re in brand or visual design, image generation models have a flat learning curve and are already producing professional-grade output when guided by someone who knows what “good” looks like.
AI in products has moved past the GPT-wrapper era, and most teams are behind the shift
Three years ago, “adding AI to your product” meant slapping a chat interface on top of an API call. Florian watched that era play out, saw the gap between expectation and reality, and thinks about it as a designer who was on the receiving end of feature requests that weren’t yet buildable.
That era is closing. The products getting it right now are ones where AI is embedded at the workflow level, where you can’t even see it’s AI, because it’s just removing friction from the thing you were already trying to do. Florian points to Intercom’s Finn as the clearest example: it started as a wrapper, took years of infrastructure work to mature, and is now genuinely capable.
The trap Florian flags is the assumption that “we have a capable model, so we can ship something capable.” The infrastructure underneath still takes time to build right. He draws a parallel to the cloud migration era of the early 2010s, when the technology existed but companies couldn’t flip overnight. The AI transition is faster, but not instantaneous.
For designers, the practical implication is that understanding AI systems at a conceptual level, what an LLM can and can’t do, what agents are actually orchestrating, how guardrails work, makes you a significantly more effective design partner on these features. You don’t need to understand how to build them. You need to understand enough to design around their real behavior, not their idealized version.
Systems thinking is the non-technical skill that unlocks AI-assisted building
One of the more useful moments in our conversation came when Florian described what actually enabled him to vibe-code things that work, not just things that look like they work. The answer wasn’t learning to code. It was systems thinking.
In Florian’s words:
“The one thing that for me unlocked all of that and enabled me to bug-fix things I would never be able to write myself is systems thinking. Understanding how the systems that AI is building for me work on a high level. Without that, I don’t know that I need a function to do this and that. I need to understand what the front end and back end are, what’s happening in my database, what happens if I need to change something there. Those concepts, if you know those, you’re good to go.”
This is a useful reframe for anyone who’s felt blocked by vibe-coding tools when things go wrong. You don’t need to write the code. You need to understand the structure well enough to describe what’s broken and what you want instead. That’s a different skill that is closer to systems design than engineering and it’s one that designers and PMs already have more of than they often realize.
His practical threshold: surface-level tools like Lovable require no visible code. If you want to go deeper into Cursor or Claude Code territory you need a working model of front-end vs. back-end, basic data flow, and what a component actually is. That’s learnable without an engineering degree.
The portfolios making Florian feel imposter syndrome share one pattern
Florian runs Open Doors, a weekly newsletter where he spotlights early-career design portfolios. He sees hundreds. And across all the different paths, university, bootcamps, self-taught, he’s identified a consistent pattern in the ones that are landing jobs at cool companies versus the ones that aren’t.
In Florian’s words:
“The people combining solid fundamentals with AI experimentation are the ones going through the roof. Those are the portfolios that make me feel imposter syndrome. The people that are struggling are the ones rushing the fundamentals, treating them as a box to check rather than a foundation to build on. Once the foundation is solid, throw AI in the mix — build something in Lovable, make it cool, post it. That’s the sequence that’s working.”
The market has shifted dramatically from a decade ago, when demand was high enough that basic portfolios could get you in the door. That’s no longer true. Hiring managers in 2026 expect typography and color to be done well, but that’s the bare minimum, not what gets you hired. Florian is direct: if you don’t meet that bar yet, the answer is to stop applying and reinvest in the foundation. Then, once it’s solid, AI amplifies what you’ve built rather than exposing what’s missing.
AI is a multiplier. Applied to strong fundamentals, it produces work that stands out. Applied before the fundamentals are there, it surfaces the gaps faster and more visibly.
AI adoption inside teams is a change-management problem, not a technology problem
Florian’s experience leading AI adoption at Juro offers a grounded view of what this actually looks like inside a real organization that’s generally open to change, had genuine friction to solve, and that still moved slowly at first.
In Florian’s words:
“Back when I started, I was sitting there with early GPT trying to write scripts and it kept failing. If I’d brought that anywhere, no one would have said ‘great, let’s have you push code now.’ But things evolved. And eventually people higher up came to me because they’d seen what I was doing. You want to be ready for that moment. And a lot of this isn’t work-unrelated. You can try it during work and call it an experiment.”
His advice for people who haven’t started yet: the window of “I’ll get to it” is closing. Four years after ChatGPT became publicly available, still not having daily usage of these tools is a real gap. Not a catastrophic one, but a meaningful one. Start with the thing that’s actually mandatory (LLM fluency), then move outward from there at a pace that doesn’t create burnout.
And for people already using these tools: the recognition that they’re imperfect is a feature, not a bug. Knowing where they fail and what they’re not good at is what makes you effective at using them where they do work.
A note from Gene
I started Uxcel because I know what it’s like to want to level up and have no accessible path to do it. When I began designing in 2010, the options were expensive courses that required you to fly to London, or bootcamps that cost five to ten grand. Neither of which was remotely accessible for most people.
What Florian said at the end of our conversation meant a lot to me: that when junior designers ask him where to start, Uxcel is now his default recommendation. Not because it’s the easiest option, but because it goes deep when depth is needed, it’s repeatable, and it doesn’t require you to bet three years or thousands of dollars on an outcome.
If you’re a designer or PM trying to close skill gaps, whether in craft, AI, or the cross-functional skills that are becoming more valuable as the triad gets closer, we’d love to help.
Sign up for Uxcel here.
Resources and tools mentioned
Tools discussed:
Lovable — AI-assisted web app builder; Florian’s recommendation for first vibe-coding experiments
Figma Make — Figma’s AI prototyping layer
Cursor — AI-powered IDE for designers working on code-heavy workflows
Claude Code — Anthropic’s agentic coding environment
Midjourney — Image generation; recommended for visual/brand designers
Intercom Fin — Referenced as a model of AI-native product evolution done right
Where to find Florian:
LinkedIn: www.linkedin.com/in/florian-boelter/
Open Doors Newsletter: www.opendoorscareers.com
Production and marketing by Uxcel. Interested in appearing on the Modern Skill Stack podcast? Reach out at partnerships@uxcel.com
What pattern do you see in your own team? Are people doubling down on fundamentals, on tools, or trying to do both at once? Drop a comment below.
Cheers,
Gene
