Here's the number that tells the real story: 89% of designers say AI makes them faster. But only 58% say it actually improves the quality of their work.
That 31-point gap, pulled from Figma's 2025 survey of designers and developers, is the most honest snapshot of where AI sits in the UX profession right now. Nearly everyone is speeding up. Barely half think the work is getting better. And yet 98% increased their AI usage in the past year anyway.
So what's going on? Designers aren't delusional. They're doing what designers have always done — picking up new tools, testing them against real work, and keeping what earns its place. The difference is that this time, the tools are reshaping the work itself.
Where AI actually delivers
AI doesn't slot neatly into one moment of the design process. It shows up across the entire workflow, but unevenly — some phases have been genuinely transformed, while others still need a human firmly at the wheel.
Discovery and research is where most designers started with AI, and it's still one of the strongest use cases. Competitive analysis that used to take days now takes hours. Drop a few competitor URLs into Gemini or ChatGPT and you'll get a structured breakdown of strengths, weaknesses, and positioning in seconds. Is it perfect? No. But it's a solid first pass that gives you something to react to instead of starting from a blank page. One Reddit commenter put it bluntly: they use Claude mainly "to unfuck PRDs from my PMs." Not glamorous, but real.
Research synthesis is another clear win. Tools like NotebookLM let you upload interview transcripts, reports, and articles, then query across them for patterns. Several practitioners described feeding in user interview transcripts and having AI pull out key themes across multiple sessions — work that used to eat entire afternoons. A designer on Reddit described their process: "I put in transcripts of user interviews and have it synthesize findings into key themes and recommendations. When I have four or five of those, I feed them in to do a full synthesis of overlapping themes."
Ideation and brainstorming gets a boost too, though a less dramatic one. FigJam's AI features can scaffold a workshop structure or brainstorming board from a single prompt. It won't replace the messy, generative energy of actual collaboration, but it removes the setup overhead — defining goals, preparing materials, structuring activities — that used to slow the process before the real thinking even began.
Prototyping is where the revolution is loudest, and most visible. This used to be the most time-consuming phase: wireframes, then high-fidelity screens, then connecting everything into a clickable prototype to demo a concept. Now you can describe a UI in plain language and get a working prototype in minutes.
Validation is still mostly human territory, but AI helps at the margins. Tools like Attention Insight simulate eye-tracking to predict where users will focus before you run any real tests. Figma plugins can audit your designs for accessibility compliance. These don't replace usability testing, but they catch the obvious issues earlier so your real tests can focus on the harder questions.
A framework for the tools
Not all AI design tools do the same thing, and it helps to think about them in three tiers based on what you're trying to accomplish.
Tier 1: Prototype generators. Tools like Vercel's v0, Stitch (from Google Labs), and Figma Make take a text prompt or a Figma file and output functional UI. V0 generates clean React components. Figma Make works inside your existing design system. These are best for rapid prototyping — getting a concept in front of stakeholders fast. They're not trying to build your production app. One designer described v0 this way: compared to tools like Lovable that tend to overdo the styling and scream "AI-generated," v0 gives you a restrained, standard-looking interface that's actually a breath of fresh air.
Tier 2: Full-stack builders. Bolt, Replit, and similar tools go further. They handle backends, APIs, authentication, payment integrations — real application logic, not just UI. A designer with no backend experience can build a working e-commerce store with Stripe payments in a single session. These are ideal for MVPs and proof-of-concepts where you need to demonstrate that something works, not just what it looks like.
Tier 3: Agentic tools. This is where things get genuinely interesting. Claude Code, Cursor, and similar agentic AI tools don't just generate code from a prompt — they work like an autonomous collaborator. They operate in your actual development environment, understand context across files, and can handle multi-step engineering tasks. A designer at Jane Street captured the shift well: "Engineers have the ability to create working proof of concept when they have an idea. Designers have to convince other people to do that for us." Agentic tools erase that gap. You can talk through a problem, get pushback on your assumptions, then build and iterate on a live interface — all in one conversation.
The practical advice here is simple: match the tool to the job. Quick prototype to test a concept? Tier 1. Working MVP with real functionality? Tier 2. Complex, iterative project that requires deep thinking? Tier 3.
The real shift: from maker to director
The tools are interesting, but the role change is the story.
Sheamus Scott Grubb, Microsoft's Principal Design Director for GenAI, quantified it. Before AI: 40 hours creating 10 design variations, then 20 hours refining them. With AI: 4 hours teaching the AI your style, 30 minutes generating 100 variations, 20 hours curating. The production phase collapsed. The curation phase — the part that requires taste, judgment, and experience — stayed exactly the same.
John Maeda, in his Design in Tech Report, frames this as a shift "from UX to AX" — User Experience to Agentic Experience. The design challenge is no longer just "How do I help someone do this?" It's increasingly "How do I help someone know whether it was done well?" When AI agents can execute tasks autonomously, the interface design problem becomes one of oversight and evaluation, not step-by-step guidance.
The Nielsen Norman Group's State of UX report put it plainly: "What isn't easy to automate is curated taste, research-informed contextual understanding, critical thinking, and careful judgment — the companies and practitioners who will thrive in this era will be the deep thinkers."
This shift is already showing up in hiring. Over 45% of design hiring managers now point to collaboration, systems thinking, and product strategy as the most critical skills — all things that require human judgment and context, not the ability to push pixels faster.
One designer on Reddit described the transformation from the inside: they went from working 60 hours a week to 5, with no one noticing, because the production work that used to fill their days now takes a fraction of the time. They're spending the surplus building their own products. That's an extreme case, but the direction is telling.
The honest caveats
This isn't a clean success story, and the parts that aren't working matter.
The trust paradox is real. 91% of UX researchers worry about hallucinations and output accuracy. 40% of designers don't trust AI-generated outputs enough to rely on them fully. The profession is in the strange position of rapidly adopting tools it doesn't fully believe in — which looks exactly like what an assistive technology looks like when it's powerful but not yet trustworthy.
Human research got more important, not less. Companies discovered that AI can generate hypotheses at scale, but it can't simulate trust. Real user research became a high-fidelity activity used to validate the mountains of AI-generated ideas. If anything, the ability to generate more concepts faster makes the validation step more critical, not less.
The junior designer squeeze is concerning. Senior and generalist roles are recovering in the job market, but entry-level positions remain scarce. When AI handles the production work that junior designers used to cut their teeth on, the profession risks losing its pipeline for developing the next generation of experienced practitioners. Jakob Nielsen has suggested an apprenticeship model may need to return.
And some designers aren't buying it at all. As one Reddit commenter put it: "I'm just wasting time trying every new bullshit autocorrect predictive LLM agent being breathlessly hyped as revolutionizing my work. None of this stuff does anything I need to do the actual work of UX." That perspective isn't the majority, but it's not fringe either. For designers deep in enterprise work with established design systems and complex stakeholder relationships, many of these tools genuinely don't address the hardest parts of the job.
Where this is heading
The designers who will thrive in the next few years aren't the ones who adopt the most tools. They're the ones who can ask the best questions, define the sharpest strategies, and know when AI output needs a human hand and when it's good enough to ship.
Figma titled their latest report "Designers Are Leaning Into the Messy Middle" — the profession is between paradigms. The old workflow of wireframes to mockups to handoff is dissolving. The new one isn't fully formed yet. The tools are powerful but imperfect. The role is expanding but uncertain.
The best move right now is the same one designers have always made with new tools: pick one up, test it against real work, keep what earns its place, and discard the rest. Just don't mistake speed for quality. The gap between those two things is where your craft still lives.
