Somewhere between "AI will replace designers" and "AI is just a tool, nothing changes," there is the actually interesting question: what does a product designer's job look like inside a company where AI is not an add-on feature but a foundational layer of how the product works?
I've been close enough to this shift — through my own work and through the teams I've advised — to have an opinion that isn't borrowed from a think-piece. The short version: the job is changing faster than most job descriptions acknowledge, and the designers who are thriving are the ones who identified the new responsibilities early and moved toward them deliberately, rather than waiting for someone to define the role for them.
What's compressing
Some parts of the traditional product design role are compressing — not disappearing, but taking less time and requiring less craft to execute at an acceptable level.
Static wireframing as the primary thinking tool is the clearest example. When AI can generate multiple layout variations from a brief description in seconds, the wireframe stops being the place where design thinking happens and becomes more of a communication artifact. The thinking still needs to happen — but it's happening earlier and in different form.
Similarly, the synthesis phase of user research — sorting observations, building affinity maps, identifying themes — is something AI handles faster and more consistently than most humans. This doesn't eliminate the research; it changes where the designer's attention goes. The job is now less about performing the synthesis and more about questioning it.
Pixel-perfect handoff is compressing too, for the obvious reason that AI-assisted code generation is making the gap between design and implementation narrower. This is genuinely good news for most designers, who never found the handoff the most intellectually interesting part of the work anyway.
What's expanding
The design surface itself is getting larger. AI-native products have interaction patterns, failure states, and user experience challenges that simply didn't exist in the previous generation of software. Someone has to design these — and the most qualified person in most organisations is the product designer, even if the job description doesn't say so yet.
Four roles I'm seeing emerge in AI-first product teams:
- AI Experience Designer. Responsible for the quality and character of AI-generated interactions: how the AI introduces itself, how it handles uncertainty, how it escalates to a human when needed, how it recovers from mistakes. This is user experience design applied to a non-deterministic system, which requires new frameworks but the same underlying instincts.
- Prompt Architect. The person who designs the instructions that shape AI behaviour within the product. A well-crafted system prompt is a design artifact — it defines the AI's persona, constraints, capabilities, and failure modes. Most companies have someone doing this work informally. In AI-first companies, it's becoming a defined role.
- AI Evaluator. When your product generates content or makes recommendations, quality assurance becomes a design problem. Someone has to define what "good" looks like, build evaluation rubrics, review edge cases, and make calls when the AI is technically correct but experientially wrong. Designers are well-positioned for this — it's essentially structured design critique applied to AI outputs.
- Trust and Transparency Designer. As AI takes on more consequential tasks — summarising medical records, making financial recommendations, driving hiring decisions — the design of how AI communicates its reasoning, its confidence, and its limitations becomes critical. Getting this wrong erodes user trust in ways that are very hard to recover from.
What stays constant
Through all of this, three things remain unchanged.
User empathy — the ability to understand what a person is actually trying to accomplish, as opposed to what they asked for — is not going anywhere. AI is very good at answering the literal question. It is much worse at knowing when the literal question isn't the real question. That gap is where product designers live.
Systems thinking — the ability to understand how individual decisions ripple through a product — is, if anything, more valuable in AI-native products, where a change to a model or a prompt can have unpredictable downstream effects across the entire user experience.
And stakeholder communication — translating between user needs, technical constraints, business goals, and design decisions — remains a fundamentally human coordination problem. AI can prepare the deck. It cannot read the room.
What to do about it
If you're a product designer who wants to stay relevant in an AI-first environment, I'd suggest picking one of the four emerging roles above and moving toward it deliberately. Not by waiting for a job posting that describes it — those are arriving slowly — but by finding the AI-powered feature in your current product that nobody has properly designed yet, and designing it.
The designers I'm most impressed by right now are not the ones who have mastered every new AI tool. They're the ones who identified that a new kind of design problem exists and applied existing instincts — user empathy, systems thinking, good judgment — to a domain that didn't have a designer working on it yet.
That's always been how the best design work gets done. It just happens to be especially true right now.