Some weeks ago, I asked senior UX expert Natalie Levy-Acosta a question I’d been sitting with for a while: What does AI actually change in the digital user-experience (DUX) design process? Not in principle, in practice. She’d spent weeks running her own experiment, stress-testing AI tools against a real UX workflow, and her findings were more precise, and more sobering, than anything I’d read in a trend report.
The short version: AI handles roughly 80% of the design process competently, the structural scaffolding, the flows, the first-pass prototype. What it cannot do is the remaining 20%: the interaction feedback that makes a user feel understood, the visual language that communicates brand without words, the creative direction that makes an experience feel like it was made by someone who cared. That 20% is not a finishing touch, it is where brand experience actually lives. And right now, most orgs are shipping the 80% and calling it done, not because they've evaluated the trade-off and accepted it, but because the tool produces something plausible enough that the question never gets asked.
Why are AI-generated interfaces starting to look the same?
The ease and efficiency of AI lures many organizations into confusing speed with quality, but they are not the same metric. The data makes this clear: Figma’s 2025 design survey found that 78% of professionals say AI tools significantly accelerate their workflows, but only 58% say AI actually improves the quality of their output. That 20-point gap matters, though not in the way you might think. Speed gains with flat quality are still a net win, the problem is not that AI makes work worse. The problem is that it makes everyone's work the same.
When every team at every company can produce a competent interface in minutes, competence stops being a differentiator. What remains, and what AI is structurally incapable of providing, is creativity, brand voice, cultural attunement, and the kind of storytelling that makes a user feel something. The Pantone Color Institute flagged this in their NYFW Spring/Summer 2026 review, noting that leading designers are actively resisting what they termed the “creeping homogenization” of AI influence. This is not a niche design-world anxiety. For senior marketers, it is a brand erosion problem hiding in plain sight.
What does the “design parity trap” actually cost your brand?
Design parity is what happens when AI-generated output becomes the default floor, and no one makes a deliberate decision to go beyond it. The interface ships, it functions, users can navigate it without confusion. But it doesn’t feel like anything in particular. It doesn’t have the personality to earn user trust, or reward attention with something genuinely unexpected.
A digital interface is not merely a utility. It is the primary surface through which users form a relationship with a brand. It does several things simultaneously: it helps users navigate and convert, yes, but it also communicates who the brand is, what it values, and whether the user belongs here. Visual language works on users the same way written language does. An interface expresses a point of view. When that point of view is indistinguishable from a competitor's, the brand has not simply lost a design point. It has lost its clearest channel for identity.
The erosion tends to follow a pattern: slow, slow, fast. In the early stages, generic interfaces feel fine, they function, they convert, no one files a complaint. But USPs blur quietly. Users stop being able to articulate why they prefer one product over another. The visual language that should be doing the work of differentiation is instead confirming sameness. By the time this registers as a business problem, declining retention, softening brand affinity, a sales team struggling to explain what makes the product distinct, reversing it is expensive and slow.
Where this ends is commoditisation. When interface design becomes a category floor rather than a competitive advantage, the only remaining differentiators are price and performance. That is a difficult position to recover from, and a more difficult one to have chosen deliberately.
How far does AI actually get you in the design process?
A standard UX workflow runs across roughly six stages: problem definition, research and synthesis, design ideation, iteration, evaluation, and documentation. What Natalie found, and what I think is genuinely useful for anyone commissioning digital work, is that AI’s value is not evenly distributed across that process:
Natalie fed a detailed brief into ChatGPT, converted it to a structured JSON prompt, and pasted it into Figma Make. Five minutes later: a fully interactive prototype with working text fields, navigable flows, a coherent visual hierarchy. It was also, by her own assessment, completely generic, indistinguishable from what any other team with access to the same tools would have produced that afternoon. Even where she had explicitly prompted for AA accessibility compliance, the output failed in multiple places when audited.
That is the real finding. Not that AI produces the same work, though it does. That it produces 80% of a product and delivers it with the confidence of a complete one.
Why does the 20% actually matter more?
The wedge I find most interesting is not between organizations that use AI and those that don’t. It is between those who use AI and maintain rigorous human creative direction, and those who allow the output to become the answer. For senior marketers, in practice it means treating AI as a sprint tool, investing the recovered time in primary research and creative direction, and asking consistently the one question no prompt can answer: does this feel like us?
AI can get you 80% of the way there in roughly 20% of the time it used to take. That is a genuine efficiency gain and it would be foolish to ignore it. But the remaining 20%, the creative direction, the brand specificity, the human judgement, now represents the entirety of what differentiates one experience from another. If you spend all of your recovered time shipping more 80%-complete work, you have not gained an advantage. You have joined the floor. The organizations that pull ahead will be the ones that used AI to move faster through the structural work, and invested what they saved into the depth, texture, and distinctiveness that no model can generate on its own.
Listen to the full conversation that sparked this piece
Natalie Flores and Yann Caloghiris go deeper on every point raised here, tool comparisons, the Figma Make experiment, accessibility blind spots, and what they actually think AI means for UX long term.