I’ve been pairing with AI a lot lately—using it to plan research, generate prototypes (case study around this one coming soon!), run workshops, and more. And while large language models have their flaws, one thing has become very clear: the quality of their output is high enough to make a lot of what we do in design feel like luxurious, wasteful exercises.
That’s not hyperbole. It’s just the direction things are going as capitalism eats itself and extracts as much efficiency and value as it can from workers. So, what’s going away soon?
Prototyping in Figma
At Shopify, everyone wanted to prototype with code (especially when they were running tests around decision-making), because code prototypes could access production data and result in more meaningful research insights.
Prototyping with code used to be a niche superpower that only a few designers had, which meant much of our testing was click-throughs in Figma with fake data. Now, with AI generating high-quality UIs, wiring up logic, and providing amazing sample data to bring it all to life, it’s becoming table stakes.
Soon, teams will be able to plug their own design systems into these models. And then, AI could produce interfaces that are not only usable but consistent with brand and interaction patterns. The difference in fidelity leads to sharper research insights—and it’s faster. Why refine lo-fi mocks when you can get a working prototype that responds to user input in minutes?
What’s especially wild is how much UI knowledge these models have available to them. Apps like Mobbin, which categorize workflows and interface patterns, make it easier for AIs to generate highly plausible UIs—often at a level equivalent to a junior or even mid-level designer. You might not need a dedicated designer to get to something shippable with a design system, and product manager or engineer with strong product sense.
A designer will still need an opinion, judgement, and maybe an initial screen (if they want to provide it to the AI to set early direction), but much of the work spent in Figma? That’s gone, and design will feel more “tangible”, like modeling clay. You’ll be able to interact with an end-to-end idea in minutes, instead of working on static screens.
Research
Qualitative analysis has always been time consuming, and and only recently has it stopped being tedious. Things have become a lot easier over the last few years thanks to tools like Dovetail, which even back in 2018 brought massive efficiencies by making research a team sport, and making tagging simpler. But who needs to tag anymore when AI can assist by transcribing, summarizing, and even cluster insights in minutes? Sure, we’ll still need human judgment to validate findings and watch out for the AI’s inevitable hallucinations (mistakes), but the legwork is easier than ever.
That said, I don’t (or don’t want to) see a world where AI will be conducting interviews. AI can help refine a script and prepare recruitment emails, but I don’t see people wanting to sit down with an AI and tell it about their time with a product and their pain points. But maybe I’m wrong – maybe video and audio generation will improve to the point where it’ll feel acceptable (and horribly dystopian) to have an hour long conversation with an AI.
At least for now though, being human and personable, building rapport, reading the room, asking the unexpected follow-up, and noticing when someone hesitates—that’s all still on us. We’re the humans.
Workshops
Workshops whose goals are to generate ideas, assumptions, and next steps may be on the way out. I’ve run multiple “five-minute workshops” with AI—posing a challenge, feeding it context, and asking it to generate assumptions, ideas, and possible next steps. The results were often on par with hour-long team sessions. Not always better, but not meaningfully worse either.
I still would’ve preferred having a workshop with my teammates, but the convenience will someday outweigh the cons when everyone is running spontaneous workshops with AI whenever they need fresh perspectives. Plus, you don’t need to check the AI’s calendar, and it never tunes out because it’s hungry.
So where does that leave us?
We still have a role to play—but it’s shifting. Instead of crafting wireframes and running sticky-note sessions, we may become more like curators, editors, or creative directors—shaping the outputs of machines and deciding what gets made, why, and how.
The design process is getting compressed, and our value will increasingly come from how well we frame problems, sense what’s missing, understand data, and guide judgment at key decision points.
Design isn’t dead. But a lot of the work we thought was design is becoming automatable. For many people, this will mean the end of their ambitions and careers in design.
I would love to end on a high note, but the best I can say for now is: I hope you now have an idea of what’s coming. I’ll be preparing a post about how to prepare by honing your judgement, practicing critical thinking, improving your data literacy, and developing your taste in visual design. The tool will not matter.