Perspective
Design × AI
AI changes who can build. It doesn't change what's worth building. The barrier to execution has dropped — which means the bar for strategic clarity has gone up. The teams that win won't be the ones moving fastest. They'll be the ones who stay clearest on what they're building, and why.
AI is changing design fast. What follows is how we're getting started.
01 — Building with AI
Designers in the code, not waiting on it
My design teams work in a Claude × VS Code × Obsidian workflow we've built and refined together. The combination lets designers work directly in code — building prototypes, shipping to production, without needing an engineer in the room for every decision. Paired with our AI-informed design system, teams can move holistically rather than in isolated handoffs.
What's working
-
·
Designers building and shipping prototypes directly
-
·
AI confidence growing across the team
-
·
Our AI design system gives everyone a shared language
-
·
Already shipping in internal tools and B2B products
What we're solving
-
·
Individual workflows are solved — team workflows aren't
-
·
Merging parallel work without breaking each other's code
-
·
Engineering collaboration workflow for merging to the customer facing app
02 — Research
AI made us more research-led, not less
The assumption is that AI replaces research. The opposite happened. By removing operational overhead, AI freed the team to focus on the parts that actually require human judgement. I led the Operations team to rebuild our research process from the ground up.
What's working
-
·
Custom AI tooling guides any team member through methodology selection, setup, screener writing, and recruitment
-
·
A Claude Code workflow handles cleaning, analysis, and synthesis
-
·
Outputs live in our internal research LLM — findable and shareable, not buried in a doc no one reads
Where we won't go
-
·
Synthetic users — not because the data can't be accurate, but because you lose the discomfort
-
·
Real research forces you to face things you didn't expect. Synthetic data lets you confirm what you already believe
-
·
That's not research. That's expensive confirmation bias
03 — Evolving the culture
Moving the whole org forward — together
At Zalando, we have the AI-obsessed, the AI-terrified, and everything in between. That gap was widening faster than organic learning could close it — and hoping people would figure it out individually wasn't a strategy.
So I organised Zalando's AI Design Day: an internal conference with external speakers, teams sharing real wins and learnings, and hands-on workshops running in parallel — from Claude 101 to building Figma plugins to getting comfortable with our AI design system.
The goal wasn't to convert the sceptics or slow down the enthusiasts. It was to move the whole org forward together. You can't afford to leave people behind — and you can't afford half your team building in a way the other half doesn't understand.
04 — Project set-up
When you move fast, alignment matters more
A misaligned team that moves slowly wastes weeks. A misaligned team that moves fast ships the wrong thing in days.
To stay aligned at speed, I've built custom AI assistants briefed on company context, ways of working, and the problem space. Drop in a problem statement and you have a sparring partner that already knows the territory — not to generate documentation, but to create shared clarity on what we're building, and what we're not.
From there, the same setup feeds into market research, conversion principles, and pattern detection. Teams get into solution space faster, with fewer false starts.
The through-line
AI removes constraints. It doesn't remove the need to think. The teams that win won't be the ones who move fastest — they'll be the ones who stay clearest about what they're building and why.