Reimagining design education in the age of AI
On rapid iteration, fresh eyes, and learning to build with LLMs
Earlier this week, I talked about the challenges facing junior designers and the erosion of traditional entry points into our industry. I ended with a call to action: It’s on us to forge new pathways and reimagine what mentorship could look like in 2024 and beyond. Enter Kelin Carolyn Zhang.
This fall, Kelin is teaching “AI Software Design Studio” to industrial design students at RISD. When the department head approached her about teaching digital product design, she had a proposal: Rather than teaching design the way it’s traditionally been taught, why not prepare students for where the industry is heading—even if we’re not quite sure where that is yet?
“Design education has historically always lagged behind industry, and for good reason—you want processes to be standardized before you can teach them,” Kelin tells me. “But right now the world is changing so quickly that even the industry is trying to figure itself out. I wanted to leapfrog this uncertainty, think about how designers should be working in five years’ time, and set students up for that future.”
The result is unlike any design class I’ve seen. These students—many of whom have never written a line of code—are shipping functional prototypes in days, not weeks. They’re learning to leverage AI while discovering its limitations. Most importantly, they’re developing a new mental model for what software design could be and how to build it.
No gods, no masters
When I ask about her students’ backgrounds, Kelin lights up. “They’re not bogged down by how product design has worked in the last decade. I’m trying to forget everything I learned and say, ‘Here’s a whole new set of technologies that nobody has fully figured out yet.’ I think they can figure it out because they’re coming at this with fresh eyes—there’s no incumbent advantage.” She makes it clear that she is trying to figure out what’s going on alongside the rest of them. “I want to just make it so clear to students that there are no gods to worship. You can figure it out, too,” she adds.
This fresh perspective is already yielding interesting results. Students are tackling ambitious projects without the usual constraints of “what’s possible.” They don’t know what should be hard, so they just... do it. When Claude (their AI assistant of choice) suggests a tech stack, they figure it out. In addition to Claude (particularly Artifacts), students use Figma, Replit, Cursor.ai, and GitHub. “Beyond that, it’s whatever is needed for the project,” says Kelin. Some students use Firebase, some use Replit’s database and deployment, one student even used MongoDB for a project. “I didn’t expect that!”
Attempts at applying any “training wheels” (like prompting Claude to write vanilla Javascript instead of React) are ignored; Students realized early on it yielded way less interesting results.
Learning by shipping
Part of that forward propulsion is the class’s aggressive ship-test-iterate cycle. Take their second project: “Bring people together.” The brief is simple—create something that connects real people, online or offline—with one twist: Test it with 100 actual users.
“I didn’t tell them how they should do it,” Kelin explains. “Thankfully, one person figured out by week two that they should deploy a website. They put QR codes around campus and they got 60 testers in a week. Then the rest of the class caught on: ‘Oh, instead of contacting testers manually, we should be deploying working websites.’ They don’t have to be perfect. The key is to hit the quantity of users.”
This rapid iteration cycle is teaching students something fundamental to software as a material: Unlike physical products, digital ones can evolve after launch. “The takeaway is realizing you can ship software and then change it later—something that’s hard to learn when you’re working with static mockups,” Kelin notes.
The AI in the room
How do the students feel about AI? It’s mixed. Kelin has been surveying them every few weeks. About half raised their hands when asked if they were scared of AI. Some enrolled to “control it,” others out of pure curiosity. But something interesting happens when you give people hands-on experience with these tools: The mystery fades, replaced by a more nuanced understanding of their capabilities and limitations.
Though a lot of the design discourse around AI tools focuses on image generators like Midjourney and Stable Diffusion, Kelin’s class mostly uses chatbot assistants to help students ship software that they couldn’t otherwise make themselves. They’re using AI to expand their skills rather than replace them.
“I constantly hear students saying ‘I followed a tutorial, or Claude told me to use this. I don’t fully know how it works, but it works,’” Kelin shares. She showed them how to ask Claude to annotate its outputs, to better understand what each line of code was doing. This creates an interesting feedback loop—students build quickly, see what works (or doesn’t), and gradually develop intuition about both the technology and their own design process.
Kelin shared a mid-term update on X after students had finished their second project, reporting that because these students were shipping first, they had different priorities from her previous design students. They cared a lot about latency and were validating ideas with prototypes and user testing before polishing the design in Figma.
These students are also bumping up against the very real technical constraints of AI tooling. Students realized chatbots gave outdated information on documentation, and were worse at making iOS apps than websites. One currently maintains three Claude subscriptions just to get around rate limits. This type of hands-on experience takes AI from an abstraction to a practical tool and part of their workflow.
The great skill gap
Of course, this approach isn’t without challenges. The biggest? Managing wildly different skill levels. “Half the class is saying ‘I want to focus more on design, I’m coding so much’ and half is saying ‘I want to be learning how to read and write code without just an LLM doing it for me, ’” Kelin admits.
This mirrors a broader challenge in design education today: How do you teach foundational skills while also pushing the boundaries of what’s possible? Kelin’s solution involves keeping the prompts open-ended to allow for different skill levels, but with specific technical requirements that force students to grapple with fundamental properties of software as a material.
For instance, their first assignment—“make a clock”—seems simple but teaches an important lesson about software’s unique ability to change over time. A clock built in software can behave differently at night versus day, respond to user input, or evolve based on data. These aren’t just technical features; they’re material properties that shape how we think about design. For students who spend their other classes building physical objects from wood, metal, and the like, Kelin’s class adds yet another dimension.
The job market question
When I ask about career prospects, Kelin is direct: “I think junior design jobs won’t exist the way we have known them. However, there’s a tremendous opportunity for junior designers to do things in a completely different way than the industry expects and very quickly surpass people who have been working in industry for a couple of years.”
The key? Not waiting for someone to create opportunities for you. “I don’t know if someone’s going to be creating those jobs for these junior designers. I think it’s going to be up to these junior designers to do it themselves,” she argues. “The job market is lagging behind all the interesting new opportunities that could come up. I’m encouraging everyone to just build the things that are interesting to them and let the rest of the industry follow in their footsteps.” (Talk about reduced latency!)
Looking ahead
What Kelin is doing at RISD isn’t just a new way to teach design—it’s a glimpse into how our entire industry might evolve. By embracing AI as a tool while maintaining focus on human creativity and problem-solving, she’s helping shape a generation of designers who aren’t just keeping up with change, but driving it.
This approach—rapid iteration, real-world testing, and fearless experimentation—might be exactly what we need to bridge the growing gap between education and industry. As traditional junior roles evolve or disappear, perhaps the answer isn’t to fight that change but to equip new designers with the tools and mindset to chart their own course.
The future of design education might look less like traditional classroom instruction and more like guided exploration—where educators create environments for students to experiment, fail fast, and discover new possibilities. It’s messy, uncertain, and exactly what we need right now.
—Carly
I love this so freakin much. And it weirdly reminds me of how I learned design and development over 20 years ago: but having an idea in mind and cobbling together things to make it happen while learning along the way. During the era of product development where Process was highly valued I really struggled to explain the way I work because I was never formally trained. I'm excited for the next generation of designers; it's the current ones I'm more worried about.
such a good and needed report/insight. Thanks for doing it!