Learn how to use AI without sacrificing quality, intentionality, and artistry. Discover 4 proven strategies for building products with soul and character.
How to Scale Quality and Creativity with AI: A Design Leader's Guide
Core Takeaways
- AI favors the probable, not the original. LLMs excel at reproducing what's popular and familiar, making it easy to create generic products that lack character and context-specific design.
- Speed creates the illusion of finished work. AI can produce polished interfaces quickly, but apparent polish often masks whether the product actually solves the problem or differentiates meaningfully.
- Intention and standards prevent "zombie UI." Defining your brand's point of view, encoding standards into your systems, and maintaining rigorous editing separates thoughtful products from thoughtless ones.
- AI amplifies human creativity when guided well. With proper inputs, iterative refinement, and cultural support for exploration, AI becomes a creative catalyst rather than a shortcut to mediocrity.
The Risk: Building Digital "Zombie Buildings"
We're in an AI building boom. Teams of three can now do what once required teams of thirty. But this mirrors the post-World War II construction explosion, when modernist design patterns were copied repeatedly without intentionality—resulting in generic, interchangeable buildings that showed no care for context or users.
The same trap awaits us in AI-driven product design. It's tempting to treat AI outputs as finished because they feel polished so quickly. Yet speed masks deeper questions: Does this actually solve the problem? Does it differentiate? Without intention, we risk creating "zombie UI"—monotonous, uncared-for interfaces that waste the time users spend looking at screens every day.
Four Strategies to Build Products with Soul
1. Have a Point of View
You must define your brand's perspective before letting AI generate solutions. Without one, AI will supply a generic default drawn from what's popular now—essentially, the past.
At Stripe, intentionality appears in every detail: color choices, writing style, what problems the product solves. This becomes your standard. When a cross-functional team debated the quality standard for an AI-generated advertisement, the insight that emerged was simple: what matters isn't how it was made, but whether it's good. Users don't care if AI created it; they care if it solves their problem and shows care.
From that conversation, the team extracted 17 improvements—softer edges, precise adjustments to visual details. This practice of meticulous craft became shorthand as "Pepsi bubbling": going one level deeper than customers see, so craft shows up in subtle ways they feel without fully seeing.
Developing this point of view requires getting good at noticing—observing what users actually need (not just what they say), studying what signals quality in products and the world around you, and building your sense of taste across art, design, and adjacent fields.
2. Encode Your Standards Into the Machine
Humans won't be in every decision. Designers won't review every interface. Agents will find and fix problems while teams sleep. Generative UI will build in real-time for users. Design systems are experiencing a renaissance, but they must evolve.
The past relied on designers filling in blanks. Now, you must make decisions explicit enough for distributed, agentic building. The object of design shifts from the screen to the system itself.
Gutenberg's printing press offers an analogy. He didn't just create 26 letters; he created 290 unique characters with different widths, abbreviations, and ligatures. This made the output both extensible and opinionated enough to feel handmade despite being machine-produced.
Modern design systems should scale intent, not just consistency. Stripe moved beyond design documentation to a CLI built on the design system itself—a "harness" that makes AI more obedient. Critically, this system includes full templates and flows, not just atomic components. It embeds what products should behave like, making it far more opinionated and coherent.
3. Refuse to Confuse Done with Good
The quality filter used to exist throughout product development. Teams had 20 ideas and could only build one, so pruning happened naturally. Now you can build 20 ideas in a week—which is powerful, but filtering must now happen after production, when saying no is harder.
The editor role becomes essential. Who in your organization views the end-to-end product and judges whether it's actually whole? This is new behavior. Experience the product as users would: Does it solve their problem? Is it coherent across the journey? Does it anticipate their needs?
A key insight from design philosophy: great art contains unexpected details—moments where you sense the maker thought of something you didn't expect—and deeper meaning that ties details together. AI is not great at this. But recognizing AI's gaps lets you fill them. As one observer noted, what offends us about AI "slop" is that details feel irrelevant—"the cup is green, but it could have been blue." An editor takes accountability for every pixel, ensuring nothing is arbitrary.
The opening animation for a recent event illustrates this. The designer iterated 56 times, refining motion, realism, and subtle life until it felt genuinely cared for. AI didn't make it one-and-done, but it opened the possibility space—allowing more complex scenes and viewpoints than would have been attempted otherwise.
4. Unleash Creativity and Artistry
AI can manufacture monotony, but it's also the greatest creative catalyst available. When everyone is building something, differentiation becomes harder and more important. Current interface paradigms—chats, charts, CLIs—aren't the epitome of modern interaction design. There's vastly more to invent.
To strengthen AI as a creative partner:
- Improve inputs. Be specific about your brand, values, and what "good" means. Don't just request a generic product; share what you believe in and source material that builds your standards. Direct AI toward different thinking modes.
- Stress outputs. Don't accept the first polished result. Push further. Use adversarial critique to challenge and refine.
- Protect exploration culturally. It's safe and easy to follow templates. It's harder and more impressive to find something unique. If you lead a team, give room to explore and protect the strange.
At Stripe, a recent cover for "Built to Grow" combined human marble artistry with AI refinement—using good human judgment to scale and perfect visual details, creating something truly stunning. This is amplification, not replacement.
The Choice Before Us
We face a choice unlike the post-war building boom. In that era, modernism's intentionality was lost to speed and copying. We can avoid that trap.
Users aren't impressed by fast production. They're impressed by solutions to real problems and clever details showing we anticipated their needs—details that reveal a brand with character and a maker who cares.
This is a moment to make a creative renaissance. Let's build products that are more powerful and show the hand and care of the maker.
Conclusion
The building boom is real, and the pace is accelerating. The difference between remarkable and forgettable comes down to one thing: intention. Your point of view. Your system. Your quality bar. Your ambition. Bring these to AI, and you don't just raise the floor—you raise the ceiling.
Original source: Raise the ceiling: how to scale intent, quality, and artistry with Al | Katie Dill (Stripe)
powered by osmu.app