There’s a conversation happening right now that goes something like this: AI is transforming design. New tools, new workflows, new possibilities. The implication is that design itself has shifted, that the arrival of generative AI has rewritten the rules. This is a 4-part series about the industry at large and the practice of designing software for people.
It hasn’t.
The principles of design are exactly what they were five years ago, and fifty years before that. Start with purpose. Understand the problem before you reach for a solution. Make deliberate choices about what to include and what to leave out. Build systems that hold together under pressure. Respect the people on the other end of what you’re making.
Design has always been the relationship between form and function, between how something works and how that working is experienced. The logic of how it’s structured. The care in how it behaves. The coherence of the system. How all of that is expressed and felt. The whole thing, not a layer applied on top.
None of that changed. What changed is that the gap between people who were actually doing those things and people who were approximating the output became impossible to ignore.
AI didn’t lower the bar for design. It lowered the cost of producing things that seem designed. That distinction matters more than most people realize.
Before AI, building a coherent product experience required real investment in how it was structured, how it behaved, how the pieces connected, how it anticipated what someone would need next. That effort acted as a filter. It took enough skill and time to produce well-functioning, well-considered work that the output itself became a rough proxy for intentionality.
That proxy is gone. AI can generate interfaces that are clean, plausible, and adequate. It can produce information architectures, interaction flows, page structures, content, and component systems that pass a first review. On the surface, the distance between “generated” and “designed” has collapsed.
What’s left, once you look beneath that surface, is whether anyone behind the work can answer the harder questions. Why does this exist? Who is it for? What are we saying no to? What do we believe about how this experience should work, and are we willing to defend that belief when it’s inconvenient?
A lot of organizations couldn’t answer those questions before AI, either. They just didn’t have to. The effort of production created enough friction that the absence of depth wasn’t visible. Now it is.
The shallowness AI is exposing didn’t start with AI. It’s the result of a decade of building products where data and metrics were the primary voice in the room, often the only voice, while design conviction was treated as optional.
That wasn’t irrational. Metrics-driven development produced measurable results. A/B testing, growth loops, engagement optimization: these frameworks delivered numbers that boards and investors understood. The problem was never measurement itself. Measurement is essential. The problem was letting measurement become the only input that mattered, letting it replace judgment rather than inform it.
When that happens, the experience layer gets quietly hollowed out. Products get more instrumented and less considered. The design layer becomes a service function: make this test variant presentable, ship this feature on time, don’t slow anything down. Nobody asks whether the product holds together as a whole, whether the information architecture still makes sense, whether the interaction patterns are consistent, whether the system rewards sustained use or just drives initial conversion. We participated in this. We optimized for what the system rewarded. The shift now is that we’re responsible for reintroducing judgment as a first-class input, not waiting for the system to do it for us.
AI arrived into that context and amplified the existing pattern. If your organization was already optimizing for speed and volume over depth and coherence, AI gave you more speed and more volume. If nobody was asking “does this experience work as a system,” AI certainly wasn’t going to ask it for you.
The result is what you might call ambient mediocrity. Left unchecked, we will produce this too. The tools don’t distinguish between teams with standards and teams without them.Output that technically functions, passes review, and ships on time, but doesn’t reward attention. Doesn’t anticipate needs. Doesn’t hold together across touch points. Doesn’t make anyone feel like someone gave a damn. That’s a conviction problem that AI made cheaper to scale.
As AI becomes infrastructure, as agents handle more execution and generation tools get embedded in every workflow, the responsibility for design is diffusing. It’s no longer contained within a design team.
Product managers are making design decisions when they define what an AI-generated experience should prioritize. Engineers are making design decisions when they architect how an agent responds, what it remembers, what it surfaces and what it buries. Marketers are making design decisions when they accept or reject generated content, when they choose which variant to scale. These are decisions about how something works, what it values, and what experience it creates.
The best people in any role were always operating across traditional boundaries. The best designers already thought like product managers and systems architects. The best PMs already cared about experience quality and long-term coherence. AI makes that cross-boundary capability the expectation rather than the exception. The neat role buckets were always a fiction, and now the fiction is harder to maintain.
When more people are making design decisions, including AI systems themselves, the need for a clear shared standard of quality goes up, not down. The counterbalance to mediocrity isn’t a better org chart or a new role definition. It’s having something to stand for, and building the infrastructure that makes that stance hold at scale regardless of who or what is doing the building.
Most of the energy in the AI conversation right now is oriented around speed and intelligence. How fast can we generate? How smart can the output be? Those are legitimate things to compete on. They’re also the ones everyone else is competing on.
There’s a parallel game organized around creativity and resonance. Creativity as the ability to see a problem differently, to frame an experience in a way nobody expected, to make a structural choice that changes how something feels to use. Resonance as the quality that makes someone come back, recommend, trust: the sense that what they’re using was built by people who understood them.
Speed and intelligence are converging. Every platform, every tool, every competitor is getting faster and smarter at roughly the same rate. The models improve, the integrations tighten, the output quality rises across the board. If you only compete on those axes, you’re in an optimization race where the ceiling is shared. The median gets better. The best doesn’t pull away. Regression to the mean.
Creativity and resonance are where differentiation survives. They require knowing what you’re trying to achieve, not what you’re trying to produce. They require the discipline to hold the line on the decisions that matter, precisely so you can move faster on everything else. They require saying no to the feature, the flow, the system that tests well in the short term but erodes trust or coherence over time.
Care is emotional. That sounds soft in a conversation dominated by benchmarks and throughput. But when everything else is commoditized, when the structure and the logic and the interface are all generated, the emotional dimension is the differentiator that remains. People know when something was built with genuine attention to how it works and how it makes them feel. They may not articulate it, but they respond to it.
If we take this seriously, this is what we need to do differently:
This is a question about what you’re building and why.
There’s a version of every product, every brand, every organization that works the way a counterfeit works. From a distance, it’s convincing. The form is right, the features are there, the functionality is close enough. But the thing that makes a great product great is the structural decisions that make it hold together over time. The interaction logic that anticipates what you need before you ask. The system design that means every piece reinforces every other piece. The philosophy that says this is how we believe this should work and then carries that through to every detail, visible and invisible.
The counterfeit gets the surface right and misses the bones. Under any sustained use, any real pressure, any real scrutiny, the difference becomes obvious.
AI has made it trivially easy to produce the equivalent of that counterfeit at scale. Entire products, brand systems, and experiences that pass the squint test. There’s a business behind that, but it’s not the kind that attracts the best people or earns lasting trust or pushes anything forward.
The other version starts with a question: What do we actually believe about how this should work, and are we willing to hold that line? In a world where execution is increasingly automated, that might be the only question that matters.
AI made the counterfeit cheap. Which makes the real thing more valuable than ever.