What UX designers actually need to survive the AI era

ELI5/TLDR

A senior product designer at Lyft argues that the survival skill for designers in the AI era isn’t another Figma plugin — it’s storytelling. AI is good at patterns, automation, and recombining what already exists; it is genuinely bad at nuance, common sense, and unfamiliar situations, and it cannot do original ideas, trust, or empathy. The thing those three weaknesses point at is one craft: telling a story about your work that makes a stakeholder root for you. The proposed move is to wrap every project pitch in a hero’s-journey arc — conflict, failed attempts, breakthrough — with the messy middle being the part AI literally cannot generate for you, because it wasn’t in the room.

The Full Story

The mood in the room

Jeong opens with the meme her audience already lives inside. Designers refresh feeds, see Claude shipped another update, see a hundred new AI design tools launched in a week, and feel a low, constant nausea about whether their job is still a job. She names it directly:

When is my job getting replaced. Is product design going to be a thing in five years. Everything I learned last month is too old now.

The point is not to dismiss the panic. Layoffs are real, AI is accelerating them, and learning the tools matters. The point is that learning more tools, on its own, isn’t the answer — because every other designer is doing the same thing.

A cleaner map of what AI actually does

Before deciding what to bet on, she splits AI’s abilities into three buckets — what it’s good at, bad at, and simply cannot do. The bucket that matters is the third one.

Good at: spotting patterns in big data, automating repetitive work (renaming Figma layers, writing boilerplate code, scaffolding prototypes from familiar UX patterns), and recombining ideas it has already seen.

Bad at: nuance (sarcasm, passive aggression, humor that depends on reading the room), common sense (the kind of thing you don’t think to verbalize), and unfamiliar scenarios (anything outside its training distribution).

Cannot do, at all: create truly original ideas, build trust through vulnerability, and feel — or generate — real empathy. Imagine the difference between ChatGPT saying “what an excellent question” for the thousandth time and a human colleague saying it once. The words are identical; the effect is not.

These three “cannots” aren’t separate weaknesses. They collapse into a single human skill — telling a story that lands.

Why storytelling, specifically

The argument has two halves. First, the cliche: facts tell, stories sell. The screens, the prototypes, the metrics — those are facts. The thing that gets people to actually use what you built is how you frame it.

Second, the structural shift. With AI tools democratizing production, anyone can ship anything. A designer can launch an app without an engineer, an engineer can design something without a designer.

It’s not about building fast or many anymore. Everyone can do it overnight while you’re sleeping.

When supply of “stuff that exists” goes infinite, the bottleneck moves to attention and persuasion. Whoever can make a stakeholder care wins.

The hero’s journey, retrofitted for a Zoom call

A good story, she says, is one where the audience roots for the protagonist. Every superhero arc — across every culture — runs the same three beats: conflict, failed attempts, breakthrough. The trick is mapping each beat to a work pitch.

Conflict = the business or user problem.
Failed attempts = previous projects that didn’t move the needle. (If there are none, substitute “cost of conflict” — how much the unsolved problem is bleeding the business.)
Breakthrough = your project, positioned as the thing that finally cracks it.

The before/after example is the cleanest part of the talk. The flat version: “this project is to reduce churn rate.” Technically correct, instantly forgettable. The hero’s-journey version walks through three quarters of consistent 15% monthly drop-off, the marketing budget being burned acquiring users who leave, the past initiatives that tried and failed, the specific insight those failures surfaced (a behavior pattern around the two-week mark), and only then the design solution that pulled churn down 8%.

The same project. One version your stakeholder half-listens to over a crying baby on Zoom. The other one they want to see succeed.

The middle is the thing AI can’t touch

The most important beat, she insists, is the ugly middle — the failed attempts, the cost, the team friction, the things you normally edit out of a work presentation because they don’t “sound great.”

Imagine there’s just a flat perfect story — there was a problem and I solved it immediately. You’re like, okay, good for you. But if you saw someone or the team struggling through something, then that’s how the audience starts rooting for you.

This is also, not coincidentally, the part AI cannot generate. AI can polish a story. It can recombine other people’s published war stories into something passable. But it wasn’t in your standup when the eng lead pushed back. It didn’t sit through the meeting where the PM misread the data. The texture of your specific mistakes is the moat.

What people latch onto, in other words, isn’t the promotion or the raise — it’s the misinterpretation, the oversight, the conflict you had to talk your way through. The polished version is forgettable; the bruised version is what gets retold.

The Korean wrap

She lands on a phrase from her first language — roughly, “to move people’s hearts.” A good story moves people. AI can simulate the move; it can’t make it. The takeaway she leaves with the room: in a few weeks the audience will forget her face and most of the slides, but if a story moved them, they’ll remember the feeling. The feeling is the durable artifact.

Key Takeaways

AI’s three “cannots” — original ideas, trust, empathy — collapse into one human skill: storytelling.
When everyone can build everything, the bottleneck moves from production to persuasion.
A good story is one where the audience roots for the protagonist. At work, the protagonist is you/your team and the audience is your stakeholder.
Hero’s-journey for work pitches: conflict (the problem) → failed attempts or cost of conflict (why it’s hard / what it’s costing) → breakthrough (your project).
The “ugly middle” is the most important and most underused beat — and the one AI structurally can’t generate, because it wasn’t in your meetings.
Polished stories from AI are flat. Original stories carry texture because the storyteller paid for it in mistakes.
People remember feelings, not slides.

Claude’s Take

This is a competent talk dressed in a slightly inflated thesis. The genuinely useful idea — that the messy middle of your project, the part you normally hide, is what makes a pitch land — is real and worth keeping. The hero’s-journey-for-work-pitches scaffold is a clean reframe of what good PMs already do; having it spelled out in three named beats makes it portable.

What gets a bit thin is the “AI can’t do storytelling” frame. AI can absolutely write you a story; the more honest claim is that AI can’t write your story, because it doesn’t have access to the specific texture of your failures. That’s the actually defensible version, and Jeong does land there by the end — but the talk spends a long time on the weaker version first.

Also, “storytelling is the survival skill for the AI era” is true in roughly the same way “communication is the survival skill” was true in 2015. The advice generalizes well past UX. That’s a feature if you’re a designer looking for permission to invest in narrative craft, and a bug if you came hoping for something specific to UX work.

Score: 6/10. Solid talk, useful framework, but a 22-minute version of an idea that fits in five. The conflict → failed attempts → breakthrough template is the keepable artifact; the rest is scaffolding.