AI Didn’t Kill Design—It Exposed It

This post is part of a series of student essays produced in collaboration with the Berkman Klein Center for Internet & Society at Harvard University. Read more in the series here.

“The design process is dead. Here’s what’s replacing it.” This is the title of an interview with Jenny Wen, the lead designer for Claude at Anthropic, that set the design world abuzz. Wen’s provocation intentionally strikes at the heart of what many product designers in the tech industry see as their core job’s purpose: to guide the direction of the products that they work on, through the design process, towards human-centered outcomes aligned with business goals. Without the design process, how do we make sure that our products are aligned with what people need?

As a designer, I can sympathize with many of my colleagues’ gut reactions to Wen’s provocation. Yet much of the discourse surrounding it fails to recognize that the adoption of AI is only making more visible the fact that the design process was never solely in pursuit of the benefit of people who use the things that we make. Power over how products are designed has always been in the hands of the companies that fund their creation, and AI makes explicit the ways that the design process is primarily used to advance company goals, rather than support the users it is meant to serve.

US courts are finally catching up to what design practitioners have always known: that the design of platforms themselves—not just the content they host—directly influences the negative behaviors that we see coming out of the use of tech, particularly on social media. The design process that shapes this form, popularized in corporate settings by IDEO’s concept of design thinking as a way of centering product decisions around human needs, has been a core value proposition of the product design discipline. Commonly used frameworks for the design process in the tech industry, such as the double diamond, start from a human-oriented place: researching the impact of a problem on real people. This research is synthesized to form a better definition of the problem, spurring the development of multiple solutions that then get tested amongst users to ultimately determine the best solution. The business value proposition is that this process guides product innovation towards solutions that people actually want and use, while the value for social good is that the process is supposed to embed real people in the creation of technology.

Contrast that design process with this scene: an anonymous user types “make me a music app with a beach vibe” into a text box and presses a button labeled “generate.” Multiple screens representing an appropriate music app instantly pop up, as if the magic wand icon accompanying the button created magic that replicated the efforts of a designer. Words describe how the user can transfer this generated design into Figma, the most common product design tool in industry, or implement it directly into code. This is the ad announcing Stitch, Google’s newest AI tool aimed at streamlining the product design process. The instant, almost magically-created, design work that strays from the user-centered approach of the design process is typical of other recently released AI-assisted design tools, such as Figma Make.

The launch of each new tool accompanies a drop in the stock prices of companies that create design software. One way to read this is that the market is signaling the future demise of the role of product designers, reckoning that their role can be replaced by AI tools that can do the same work faster and more seamlessly with engineering. The argument that Wen makes for this new way of working aligns with that: working hand in hand with engineers to quickly iterate, with AI-empowered engineering implementation driving product direction.

However, during the seven years I spent working in the tech industry as product designer for emerging tech firms before the rise of generative AI design tools, I found that instead of spending most of my time experimenting with new solutions to help connect people better through technology, I spent more time convincing others that doing good by the people using our products was aligned with business goals. On one product I worked on, it took months of self-driven user research and internal advocacy work to successfully advocate for women to be considered an integral part of the product’s target audience. On another product, I was unsuccessful in persuading management to find alternatives to lootbox-style mechanics and eventually the output of my work was an extractive gamification system that powered the product’s social dynamics. When I attempted to experiment with interactions and mechanics, it would frequently conflict with product roadmaps that prioritized monetization features.

As a designer who values doing good by the people that I design products for, those experiences made clear to me that the incentives of the tech industry often were at odds with how tech companies branded themselves as doing good in the world. Kentaro Toyama popularized the idea that technology by itself does not solve social problems, but that it only amplifies human intent and capacity. That is, technology amplifies existing power structures and incentives, instead of acting as a solution to them. In line with this theory, I believe that AI has simply amplified the incentives that currently exist in the tech industry, and the benefits that AI companies tout as evidence of their ability to change the world only reveal the values that underpin how decisions are made within the tech industry.

Instead of waiting for designers to iterate on interfaces before engineers implement them, engineers themselves can spin up multiple designs on their own to test out ideas, shifting decision making power towards those who rush to implement. Designers get caught up in the demand for quicker solutions and learn to spin up prototypes in order to remain part of the decision making process. Where speed is prioritized because it provides companies with clear advantages—first to market, quicker time to release features that a competitor does not have—AI provides an answer.

However, these shorter development cycles also lead to shrinking windows of agency for designers to intervene in decisions that may serve company goals but run counter to what they believe is best for users, because there is simply less time to raise concerns or do the advocacy work needed to convince management of a shift in direction. The problem becomes compounded as less labor and company resources are needed to ship a feature. Because of the lower cost of shipping a feature, the default answer to a concern raised becomes “let’s see how it goes—we can easily change it later if it doesn’t work.”

Wen’s approach to ship fast and iterate publicly “building trust through speed,” is representative of those values and it is the newest embodiment of the old Facebook motto that the tech industry has baked into its culture: “Move fast and break things.” When designers, the ones tasked with advocating for users, lose the agency to intervene in product decisions, the consequences are passed onto the millions of users that rely on those products. Products become less responsive to their needs and more extractive in their design.

I see this already in my work. Before design and product teams agree on important requirements or understand what users really need, engineering has already spun up a prototype that meets their understanding of the problem. These prototypes lead the conversations about the product, as it is easier for people to respond to existing artifacts that they can interact with, even when they conflict with the actual user’s needs. Worse, I have been presented with engineering builds that embed assumptions that conflict with user needs into the underlying system architecture, making them much harder to change without a significant overhaul of the app itself.

While US courts are only recently grappling with the ability to hold tech companies accountable for their design choices on social media platforms, the product design field is already using AI to accelerate the potential to create even more design paradigms that negatively impact users that courts have yet to contend with. I do not see this as inherently a problem with AI—it is simply amplifying the existing problems in the practice of product design. AI is not going to make companies value design if they did not value it to begin with, nor will it make companies care about user edge cases or the negative impact of their product on people’s lives more than it did before. That would require a shift in the tech industry towards valuing the importance of thoughtful design. More importantly, it would require a paradigm shift towards an industry-wide questioning of the values of speed, productivity, and profit that underpin how the tech industry builds and evaluates its products. More fundamentally, it would require a reconsideration of who should have the power to shape the systems that millions of people rely on.