Stripe Head of Design Katie Dill Breaks Down Their New Website

Why does it matter? Because most websites are manifestos in disguise — and yours is probably lying.

Katie dill spent over a year redesigning stripe.com — a site that was already six years old and still looked modern. What she uncovered isn't really about web design. It's about what happens when a company confuses "good enough" with "done," and why the visual choices on a public-facing product are actually claims about your engineering culture.

• A homepage isn't decoration — it's a trust signal that tells customers whether to believe your core product is also carefully built
• AI raises the floor of design to a reliable 7/10, but teams that stop there are training themselves to accept mediocrity everywhere
• The bento-plus-modal format beat every alternative in user testing because it keeps visitors in browse mode instead of forcing commitment
• Design systems are about to evolve from style guides into AI-managed infrastructure — and companies with weak systems will feel it fast

Your website is your manifesto — and every sloppy pixel signals that your product is sloppy too

"If you see the care that goes here," Katie says, "then you're right to assume that we also put that care behind the scenes in the way that we move money and the way that we protect a company's information."

For a financial infrastructure company, a muddy homepage isn't just an aesthetic miss — it undermines the entire sales argument. Every typography choice, every gradient decision, every animation tuned half a degree too loose tells an enterprise buyer something about what they can expect when you're handling their transactions at 3am.

This is what pushed the redesign even though the old site was holding up fine visually. The business had outgrown the story. Six years ago, payments was the product. Today, that's one item in a suite serving multinational enterprises, AI startups on usage-based billing, platforms like a market equivalent of a marketplace. The old homepage was adding sections like a hoarder adds rooms — coherent locally, incoherent as a whole.

Patrick asked the team a simple question during planning: what is the actual point of a website? Katie's answer: it's your manifesto whether you call it that or not. The clothes you choose to wear for the next six years. That framing — high-stakes, long-horizon, deeply identity-linked — is what justified a year of iteration before shipping anything.

The S&P rally is real — but it might be inflation in disguise

Skip that — wrong pod. Here's the real one:

The gravitational pull of every design process is toward mediocrity, and the only defense is explicit, daily resistance

"The gravitational pull is to mediocrity. It is just so easy to accept good enough." Katie says this plainly, without hedging. And then she does the math that makes it uncomfortable: imagine making the 'good enough' call every day, on every decision, across an entire company. That's not a hypothetical — that's most companies.

The trap isn't one big capitulation. It's the composite of a thousand small ones. A data visualization that's pretty good. An animation that's almost right. A wave gradient that cleared the bar but didn't clear the bar you actually wanted. Each individual choice seems defensible. The aggregate is a product that feels definitively mediocre, and nobody can point to why.

Katie's antidote isn't a process — it's a posture. Don't let effort already expended become a reason to ship something subpar. The team worked hard on it is not a quality argument. "How happy is the team going to be if the product is just meh?" The counter-pressure has to come from the other side: what is the cost of shipping this? Will it erode trust? Does it feel like care, or does it feel like someone ran out of time?

Her actual instruction to her team: "Fight the gravitational pull, the mediocrity, and do not leave well enough alone."

AI gets you to a 7/10 in minutes — teams that mistake that for done will ship slop

Twenty concepts where you used to explore two. That's the real acceleration AI brings to design exploration, and it's not nothing. But Katie flags a specific psychological trap that comes with it: the ease of generation makes the output feel earned.

You type a prompt, something comes back, it looks competent, and there's a pull to call it done. "Don't be wooed by just how easy that was to achieve, but instead ask yourself: but is this really great?"

The customer images on the new site are the clearest example. The brief looked simple — a parallelogram, some brand colors, a real-feeling photo. AI could generate 50 versions fast. But the team's internal critique sheets were dense with specifics: the arm isn't anatomically right, the hand is missing, the shadow doesn't fall correctly. Each issue invisible at a glance, viscerally wrong on second look.

"It doesn't replace craft. It doesn't replace taste. It doesn't replace the attention to detail to ensure that you're getting each of those things right."

The actual workflow: AI accelerates exploration and prototyping — how should these accordions move? what does this interaction feel like? — and saves engineers hours on variants that won't survive user testing anyway. But a human with taste has to be the final gate. Speed in service of iteration is the win. Speed as a substitute for judgment is how you ship slop at scale.

The bento-plus-modal beat every other format because it respects how people actually browse

The old site had the classic accumulation problem: new product, add a section. New customer segment, add a section. Eventually you have a page that requires sustained scrolling commitment before a visitor understands what you actually sell.

The team tested alternatives. A full-screen collapsed view — too text-heavy, not enough showing versus telling. A section-by-section scroll format — too slow, too much asking visitors to track a narrative. An accordion — taken to user research, confirmed what most designers already know but rarely act on: "It was not a quick way for people to really digest a lot at once because it requires effort." Even a small click, when you're in browse mode, breaks the spell.

The bento won because it keeps users in what Katie calls "lean back" mode. Six tiles, each with minimal text and motion that signals interactivity. Want more? A modal opens directly on the page — no navigation, no context switch, no back button required. "The idea that you're going to put your most important messages a click away, it's going to be tough."

The insight generalizes: information architecture isn't just a UX preference. It's a conversion variable. Every format decision either preserves the visitor's browse posture or breaks it.

The GDP counter is the rare trust signal that answers the right question for enterprise buyers

78% of the Forbes AI 50 use stripe. That's a logo wall in sentence form — useful, but it answers "do others trust you?" not "can you handle my scale?"

The live GDP counter does something different. It answers in real time: here is the fraction of global economic activity that runs through our infrastructure right now. For an enterprise buyer whose core fear is "will this thing fall over when we actually need it," a number that keeps ticking is more viscerally convincing than any case study.

"It matters a lot to them," Katie says, "whether or not that stripe is dependable, reliable, trustworthy, and can operate at their scale."

Almost no company can replicate this specific metric — which is precisely the point. The design principle it surfaces is worth extracting: find the single number that proves you operate at the scale your target customer cares about, and put it at the top of the page. Not buried in a case study. Not footnoted in a press release. Right there, in the first scroll.

Walking the store every Friday — with engineers, PMs, and designers in the same room — is the most underrated quality practice in product

Products built by separate teams drift into incoherence invisibly. One team ships a new dashboard. Another updates the billing flow. Nobody updated the navigation in between. "It's in your home when you've updated the dining room, but none of the other rooms. And all of a sudden, the light switches don't match."

Stripe's fix: a standing Friday ritual where the founders walk the live product in front of the whole company. Not a demo of what's coming — a walk of what's actually shipped, right now, as a user would encounter it.

But the format matters as much as the cadence. Katie specifically wants an engineer, a product leader, and a data scientist in the room at the same time. Different disciplines surface different failures. The engineer notices a performance issue. The data scientist notices a drop-off point. The designer notices an animation that's slightly wrong. None of them see what the others see — which means a room of only designers, or only engineers, leaves whole categories of problems invisible.

The practice isn't exotic. It's a scheduled, recurring, cross-functional review of the live experience. Most teams don't do it. The ones that do catch fragmentation before users do.

Design systems are about to become AI-managed quality infrastructure — and weak ones will collapse under the pressure

The trend everyone's watching is democratized code — more people inside companies shipping features, updating pages, modifying products without going through a design review. Katie sees a second-order consequence most teams aren't ready for: if everyone can ship, the only scalable quality control is a system that enforces standards automatically.

She sketched an interaction on paper, fed it to an AI tool, and watched it assemble components from the existing design system into a working prototype. The floor of quality — for anyone in the company — becomes whatever the design system is. Then the next question: can the AI tool identify gaps in the system itself and help expand it for new use cases?

Companies that have invested in rigorous, well-documented design systems are quietly building AI leverage. Companies that haven't are about to discover that AI-generated code proliferates faster than human reviewers can check it.

Where this is all heading: taste becomes the scarce resource

As AI compresses the time between idea and shipped product toward zero, the bottleneck shifts from execution to judgment. The teams that win won't be the ones who move fastest — they'll be the ones with enough taste to know when fast is good enough and when it isn't.

The design system insight points somewhere specific: the companies building rigorous quality infrastructure now are encoding taste into tooling. Everyone else will be debugging drift after the fact.

The floor rises. The ceiling stays exactly where craft and attention put it.

---

Topics: design, product design, stripe, website redesign, AI in design, design systems, UX, brand, prototyping, design process