Key Takeaways
- After World War II, builders threw up sterile, copy-paste concrete blocks that turned cities into uniform grids of "zombie buildings." Software faces the same fate right now as automated code tools flood the web with identical chat boxes, simple charts, and command lines.
- In 1850, art critic John Ruskin praised Gothic architecture because no two Gothic structures looked alike: every mason's personal care, quirk, and judgment stayed visible on the stone.
- Machine models optimize for statistical averages. If you ask automated tools to build a screen without clear constraints, they will output the most generic version of that interface.
- Fast generation bypasses traditional pre-build design gates, which means product leaders must shift from planning specs to acting as aggressive post-build editors.
- Teams can avoid bland interfaces by applying Katie Dill's 4 Recommendations for Building Products with Care and Soul Using AI.
Katie Dill's 4 Recommendations for Building Products with Care and Soul Using AI
1. Have a Point of View
Define your brand, who you want to be to your users, and what they care about. If you don't supply a point of view, AI will provide a generic one based on past popularity. Hone your point of view by getting good at noticing what signals good, great, and mediocre in the world around you.
2. Encode Your Standards into the Machine
Make the design system itself the primary object of design. Build opinionated frameworks (such as CLIs with full flows and templates rather than basic atomic components) that embed standards directly into the means of production so agentic and distributed builds scale intent, not just consistency.
3. Refuse to Confuse Done with Good
Overcome the temptation of premature completion (the burrito dilemma). Because traditional pre-build quality filters are bypassed by fast generative tools, adopt the role of a post-build editor who experiences the product end-to-end like a user, pushes through dozens of iterations, and takes accountability for every detail.
4. Unleash Creativity and Artistry
Treat AI as a creative partner rather than a shortcut to cookie-cutter patterns. Improve inputs with brand-specific prompts and source material, stress outputs using adversarial agents, invent new interface modalities beyond basic chat/CLI, and protect the strange by reinvesting efficiency gains into ambitious craft.
When This Works (and When It Doesn't)
This approach works for product, design, and engineering teams building software with generative tools who want to stand out instead of shipping lifeless, uncared-for screens. When your market is crowded, the speed of shipping is no longer an advantage on its own. Everyone can prompt a working prototype in ten minutes. Care and taste become your only real moats.
It fails when founders apply craft perfectionism to throwaway internal scripts or early validation experiments that only need to test a raw database query. If your user only cares about a fast back-end transaction, do not spend three weeks custom-styling an interface. Reserve this high-effort editorial eye for customer-facing flows that define your product's core identity.
What to Do With This
Take the most critical user journey in your product this Friday, such as your onboarding or checkout flow. Generate three fresh variants using an automated tool, but do not stop at the first working output.
Put on the hat of a ruthless post-build editor. Walk through the full screen flow on a live device, click every button, and spot where the machine settled for a cookie-cutter chat prompt or an empty table state. Reinvest the four hours your engineers saved on boilerplate code into building custom micro-interactions, sharper brand copy, and unexpected visual feedback that make the software feel responsive and distinct.