From interface design to system design

Over the past year, everyone seems to be talking about how designers need to evolve. Over the last eight months, I've been experiencing that shift firsthand. A year ago, my work was what most people would consider traditional UX design: mapping user flows, creating wireframes, designing high-fidelity mockups, and handing them off to frontend engineers. Today, I still do all of those things. But I also write code, push PRs, and design agent workflows. What's interesting is that this wasn't a conscious decision to become an engineer. Instead, it was a response to the way AI products are changing the nature of design itself. As users stop following predefined flows and product experiences become increasingly shaped by models, agents, and system behavior, designers can no longer operate solely at the interface layer.To design the experience, you increasingly need to understand the system behind it.Code is only the entry point.The deeper shift is that designers are beginning to move from Interface Design toward System Design.

UX emerged during a very specific era of software. It solved an important problem and played a critical role in the growth of the software industry. But it was also built on a set of assumptions that AI is now beginning to challenge. In the early days of GUI, programmers were responsible for everything: backend logic, frontend implementation, and the interface itself. As software became more complex, specialization emerged. Engineering split into frontend and backend disciplines, while design evolved into its own field focused on understanding users and shaping their interactions with technology. This division of labor worked remarkably well. Designers focused on user needs, usability, and reducing friction. Engineers focused on building and maintaining the underlying systems. A key reason this model worked was that traditional software was largely deterministic. User journeys were predefined. Product behavior was predictable. Designers could create great experiences without needing to deeply understand the systems behind them. AI products fundamentally change that assumption. Users no longer follow carefully designed flows. They can ask anything, express any intent, and interact with products in ways that are impossible to fully anticipate. At the same time, product behavior is no longer entirely defined by the designer. Increasingly, it is shaped by models and the decisions made by the system itself. As a result, design questions begin to shift. Instead of asking:"What should the user click next?" We start asking: "Which tool should the system use?" "How do we help users understand what's happening behind the scenes?" These are still user experience questions. But they can no longer be solved through interface design alone. The complexity that was once hidden behind the interface is now becoming part of the design problem itself.

AI didn't create an entirely new problem. What it did was make an existing one impossible to ignore. For a long time, there has been a growing distance between designers and the systems behind the products they create. In the era of deterministic software, that distance was manageable. Today, as product behavior becomes increasingly shaped by models, understanding the system itself is becoming an essential part of the design process. This means that learning AI alone is not enough.It also requires a deeper understanding of how software systems work.After all, AI is not a completely separate world. It is better understood as a new layer of abstraction built on top of existing computer systems.

Fortunately, I think there has never been a better time to learn these things. AI is not only changing the way software is built; it is also changing the way we learn to build it. Tools like Cursor, ChatGPT, and Claude have dramatically lowered the barrier to entry. Concepts that once required years of experience to approach are now far more accessible than they were even a few years ago.

Over the past eight months, I've been gradually moving closer to the product itself. I began reading and building components, understanding APIs, and participating in interactions that spanned both the frontend and backend. What surprised me was that the more I got involved, the more I realized how incomplete my understanding of the system actually was. At the same time, I found myself spending more time learning about AI itself. Many of the most important product challenges were no longer UI problems. They were questions about model capabilities, context management, tool invocation strategies, or the design of the system as a whole. Looking back, the most important change over these eight months wasn't that I learned how to write code. Code was just the entry point. The deeper shift was that I stopped focusing solely on what users see and started asking why the system produces the outcomes that it does. I began moving beyond designing interfaces and toward understanding the system itself. But this doesn't mean I need to become engineer. Understanding the system is ultimately a way to better understand the user. Once designers begin to understand both sides, they gain a more complete perspective. They can see not only what users are experiencing, but also why the system behaves the way it does. And that perspective makes it possible to design solutions that bridge the gap between user needs and system capabilities.

From this perspective, the best designers of the future will still be the ones who understand users the most. What is changing is not the goal of design, but the scope of what designers need to understand. And it is this ability to bridge user needs and system capabilities that will allow designers to continue playing a central role in building great products in the age of AI.