Artificial intelligence is rapidly becoming one of the defining technologies of our age.

Every week brings another announcement promising transformation: faster workflows, automated creativity, infinite scalability, personalized everything. Entire industries are reorganizing themselves around the assumption that AI will fundamentally reshape how products, services, and organizations are designed.

In many ways, this assumption is correct.

AI is already demonstrating extraordinary capability in areas such as synthesis, pattern recognition, optimization, summarization, and automation. It can generate code, draft interfaces, write marketing copy, analyze research, produce visual assets, and dramatically accelerate production pipelines. For organizations focused on efficiency and scale, the appeal is obvious—and on a personal level, it's "unlocked" me to get things out of my head and into a prototype faster than ever before.

But beneath the excitement lies a deeper question. One that's less discussed, and perhaps more important. What if AI is exceptionally good at extending existing ideas, but fundamentally limited in its ability to redefine them?

This distinction matters because not all innovation is the same. Much of what organizations describe as innovation is, in reality, incremental optimization. Products become faster, cheaper, smoother, or more feature-rich. Existing systems are refined and improved. There's undeniable value in this kind of progress, and AI is likely to accelerate it dramatically.

But radical innovation is something different entirely. It doesn't simply improve an existing frame. It changes the frame itself.

Design theorists such as Roberto Verganti have argued that the most transformative innovations aren't technological breakthroughs alone, but "innovations of meaning." These innovations fundamentally alter how people understand a product, service, or experience. Similarly, Kees Dorst describes "frame innovation" as the process of redefining the problem itself rather than merely solving it more efficiently.

This distinction feels increasingly important in the age of AI because many current applications of AI operate within existing frames. They optimize known variables. They extend patterns already present within their training data. They recombine established ideas in sophisticated and often useful ways.

But radical innovation frequently emerges from questioning assumptions that nobody realized were assumptions in the first place.

In the mid twentieth century, a number of office buildings faced persistent complaints about slow elevators. Engineers initially approached the problem in the obvious way: increase elevator speed, redesign the lift systems, or install additional elevators. All of these solutions were technically expensive and operationally complex.

Instead, one building owner reframed the problem entirely. Mirrors were installed beside the elevators, and the complaints largely disappeared.

The issue had never really been elevator speed. It was the psychological experience of waiting. People were bored. Impatient. Uncomfortable standing idle. The mirrors gave occupants something to do. They altered the perceived experience of time—helped by the fact that many of us like looking at ourselves in the mirror.

This solution wasn't an engineering optimization or a process improvement. It was a reframing of the problem itself, and of the inherent limitations of the people defining the problem.

The distinction is subtle but critical. Engineers were asking: "How do we make the elevators faster?" The reframed question became: "How do we make waiting feel shorter?"

This is where many contemporary conversations about AI begin to feel incomplete.

AI is extraordinarily capable at answering questions within an existing frame. If the problem is clearly defined, measurable, and pattern-based, AI can often outperform humans in speed and consistency. In design contexts, this means AI is increasingly effective at generating interface variations, summarizing research findings, producing design systems, optimizing workflows, and automating repetitive tasks.

But the challenge is that many of the world's most meaningful problems are poorly framed to begin with.

Healthcare systems don't merely suffer from inefficiency. Education systems aren't simply content delivery problems. Loneliness can't be solved solely through communication tools. Urban congestion isn't always a transport issue. Often, the underlying problem is social, cultural, emotional, or behavioral.

The difficulty lies not in solving the problem faster, but in understanding what the real problem actually is.

This is why design, at its best, has never been purely about aesthetics or usability. Good design is fundamentally interpretive. It's concerned with understanding human experience, surfacing hidden tensions, and redefining problems in ways that create new possibilities.

Historically, many of the most transformative products emerged from this kind of reframing. The smartphone wasn't merely a better phone—it reframed the phone as a general-purpose computing platform. Streaming services didn't simply improve video rental logistics; they changed expectations around ownership and access. Social media didn't merely digitize communication; it transformed identity, social signaling, and participation itself.

These innovations weren't simply better executions of existing ideas. They altered the meaning of the product category—like Bang & Olufsen, who redefined what sound equipment could be, pushing beyond boring hi-fi units and making the equipment pieces of art and furniture.

AI, however, largely operates through probabilistic synthesis of prior human outputs. It predicts likely continuations. It identifies patterns. It generates responses based on learned relationships between concepts. This makes it exceptionally powerful at extension and iteration.

But frame innovation often requires stepping outside the logic of the existing system altogether. It requires interpretation, judgment, and cultural awareness. It requires understanding contradictions in human behavior that may not yet exist clearly in data. Most importantly, it requires asking questions that nobody thought to ask.

This isn't an argument against AI. In fact, AI will almost certainly become one of the most powerful creative and operational tools ever developed. Designers, researchers, engineers, and strategists who ignore it will likely place themselves at a disadvantage.

The issue isn't whether AI is useful. The issue is how narrowly we define innovation itself.

If organizations begin to equate innovation purely with optimization, acceleration, and efficiency, then AI may unintentionally reinforce existing systems rather than challenge them. We risk producing increasingly sophisticated versions of the same ideas, the same business models, and the same assumptions.

More content, more features, more automation, more speed—but not necessarily more meaning.

This matters because many contemporary technology products already suffer from a kind of optimization obsession. Metrics become proxies for value. Engagement replaces fulfillment. Efficiency becomes more important than experience. Entire industries become focused on removing friction without questioning whether friction itself sometimes serves a meaningful human purpose.

I have a mechanical watch. It must be wound to operate, and I often have to re-set the time when it winds down. There are more efficient, cheaper watches that don't have these "limitations"—yet I (and many others) find a mechanical watch far more desirable than a battery-powered quartz watch. Its purpose is greater than just a mechanism to tell the time. It's a piece of art, jewelry—and it's a joy to wind it.

Similarly, the manual transmission in a car has largely been replaced by automatic gearboxes that change gears faster through a paddle shift. On paper, these gearboxes perform better than a manual. But in many cases, I find the experience soulless compared to the physical movements of changing gears with a clutch pedal and gear stick.

Not every inefficiency is a problem. Not every delay is bad. Not every interaction should be optimized into invisibility.

The elevator mirror example demonstrates this perfectly. The most effective solution wasn't technological acceleration. It was understanding human psychology.

As AI increasingly automates production, synthesis, and execution, uniquely human capabilities may become even more valuable. Not because humans are inherently superior, but because human creativity is deeply entangled with lived experience, emotion, ambiguity, culture, ethics, and interpretation. Humans don't simply process information. They assign meaning and value to it.

And meaning is often where radical innovation begins.

The organizations that succeed may not be those who simply deploy AI the fastest, but those who use AI while still cultivating human judgment, curiosity, interpretation, and critical reflection. The real opportunity may lie in combining machine capability with human reframing.

AI can help us explore solutions at unprecedented speed. But humans must still decide which questions are worth asking—and what problems are worth solving.

Because ultimately, the most transformative innovations rarely emerge from optimizing the world as it is. They emerge from imagining the world differently.