Elin Forsberg
“A design system used to define pixels. Now it has to define how a screen behaves when it's wrong.”
- Martin Knapic, Product MoatElin Forsberg has spent fourteen years building design systems, which used to mean pixels and spacing tokens and now, in her words, means teaching a screen how to behave when it's wrong. She leads product design at Anchorpoint in Stockholm and was one of the first people in this series to make the case that generative interfaces need entirely new component patterns, not just new colors.
Identity & Background
- Name
- Elin Forsberg
- Current role
- Head of Product Design
- Current company
- Anchorpoint
- Location
- Stockholm, Sweden
- Years in product
- 14 years
Your story: how did you end up in product?
I trained as an industrial designer, actually - physical objects, tolerances, manufacturing constraints. Screens felt looser and less honest to me at first, until I realized interaction design has its own physics; you just have to learn where the edges are.
Core PM Philosophy
How do you define product success?
Whether a team can ship a new pattern without asking design for permission. A design system succeeds when it makes the good choice the easy one, and design becomes a partner rather than a gatekeeper.
What's your decision-making philosophy?
I ask what the user would guess happens next, then test that guess with five people before we build anything. If more than one is surprised, we go back to the drawing board, however good it looks.
How do you manage stakeholders?
Engineers and PMs get a seat in system decisions early, and I explain trade-offs in terms of their constraints - build time, accessibility risk, maintenance - not design principles.
Product & Decision-Making
What product inspires you?
Teenage Engineering's hardware. Every object has a strong point of view, a limited vocabulary, and deep respect for constraint - exactly what I want a design system to feel like.
AI & the Future
How is AI changing product management?
Static comps stopped being the deliverable for anything generative - I now design ranges of acceptable output and the states around them, and I prototype directly against real model calls instead of faking data in Figma, because fake data hides exactly the failure modes that matter.
Which AI product do you use and respect?
Claude for structuring critique and generating variations to react against, and Figma's AI features for tedious layer cleanup. I never use it for the actual concept - that still has to come from a person who's sat with the problem.
What's one thing about AI everyone gets wrong?
Treating the chat box as the default interface for everything. Half the time a form, a slider, or a plain button gets the job done faster and with less ambiguity than a prompt ever will.
Beliefs & Opinions
What's your unpopular product opinion?
Design debt in AI features specifically - the confidence-signal patterns and fallback states we're shipping today are mostly one-offs, invented per team, with no shared language yet. I think the next real design-system work is standardizing that vocabulary before every product invents its own, incompatible version.
What makes a great product?
Legibility of intent. Whether it's a dialog box or a generative agent, the user should be able to guess what happens next before they click. That principle predates computers and it'll outlast this AI cycle too.
Company & Product
What does your product solve?
Financial planning software for independent advisors - the interface an advisor uses in front of a client to model scenarios live, in the room, rather than presenting a static PDF after the fact.
What's the market opportunity?
Our ICP is an independent advisor or small RIA, not a wirehouse - someone who chose independence and needs software that doesn't feel like it was designed for a 10,000-person firm. Against the big platforms, we're deliberately less powerful and much faster to actually use live with a client.
More from the conversation
What's a hobby that shapes how you think about design?
Ceramics, on a wheel, most Sunday mornings. Nothing teaches you about constraint and material behavior like clay that collapses the moment you push it past what it can hold - it's the most direct feedback loop I have all week.
Was there a point you nearly left the field?
A stretch doing pure visual design with zero research access, just executing someone else's decisions. I moved to a smaller company specifically to get research access back, even for less title. It was the right trade.
How big is your design org, and where does design sit in the company?
Nine designers inside a product and engineering org of about 90, at a company of 250. Design reports alongside product, not under it, which I fought for specifically because design-system decisions need equal weight with roadmap decisions, not just execution power.
What's the biggest gap in most design systems today?
Almost none of them have patterns for partial confidence, streaming content, or graceful wrongness. We spent a decade perfecting empty states and loading states. Now we need 'plausible but unverified' states, and most component libraries have nothing for that.
How do you design trust into an AI feature?
Show your work in the smallest possible unit - a source, a confidence signal, an easy undo - rather than a giant disclaimer banner. People calibrate trust from small, repeated cues, not from a paragraph they'll skip.
Advice for designers moving into AI product work?
Sit in on the eval reviews, not just the design crits. Watching a hundred real model outputs teaches you more about the actual failure modes than any Figma prototype will.
Enjoyed this conversation?
A new interview with a product person lands every week. Get it in your inbox.