The Part AI Still Can’t Do (And Why That’s Where Designers Matter Most)

AI generates screens in seconds. It still can't tell you which problem to solve. That's where designers matter most.

AI can generate ten screens in thirty seconds. It can produce responsive layouts, suggest color systems, and export clean code. The AI UX design future is already here in many ways. But there’s a part of the design process that AI hasn’t touched, and I don’t think it will anytime soon.

The concept phase is not a production problem

Most AI design tools solve a production problem. They make it faster to go from idea to artifact. Describe a screen, get a screen. Upload a sketch, get a polished mockup. That’s valuable. But the concept phase of design isn’t primarily about producing artifacts. It’s about figuring out what to produce.

When I researched this topic in 2023, the literature was clear: the concept phase involves understanding user needs, identifying the right problem, generating and evaluating ideas, and making decisions about direction. Tools can support that process. They can’t replace the thinking.

Three years later, the tools are dramatically better at the production side. But the thinking side? That’s still entirely human.

What AI can’t do in the concept phase

AI can’t sit in a room with a client and sense that what they’re asking for isn’t actually what they need. It can’t notice that a user’s frustration with a checkout flow is really about trust, not about the number of steps. It can’t look at research data and decide that the real opportunity isn’t in the data at all, but in something nobody mentioned.

These are judgment calls. They require empathy, experience, and the ability to hold ambiguity long enough to find something interesting in it. That’s what designers bring to the concept phase, and it’s the part that matters most for whether the final product is actually useful.

My thesis used Arvola’s framework to analyze how AI tools support different aspects of usability. The technical and practical dimensions have seen enormous progress. The communicative dimension is improving. But the organizational, ethical, and deeply human aspects of design work remain firmly in the designer’s domain.

The AI UX design future is about asking better questions

Here’s what I think the next few years look like. The production gap between designers and non-designers will continue to shrink. Product managers, developers, and founders will generate perfectly acceptable screens with AI tools. The visual output will be good enough for most contexts.

That means the value of a designer shifts. It moves further up the process, into the concept phase, into the strategic decisions that determine whether a product solves a real problem for real people. The designers who thrive will be the ones who are excellent at the parts AI can’t automate: research, synthesis, judgment, and the courage to push back on a brief that’s pointing in the wrong direction.

This isn’t a threat to designers. It’s a clarification of what design actually is. It was never about pushing pixels. The pixels are automated now. Design is about making the right decisions about what to build and why.

What I’m watching for

A few things I’m paying attention to as this space evolves.

First, how AI tools handle the ethical dimension. Training data transparency, output bias, and data privacy aren’t solved problems. They’re getting more complex as more proprietary design work flows through these tools.

Second, how design education adapts. My thesis specifically called out the need for AI competence in UX education. That’s still a gap. Most design programs teach tools and visual skills. Fewer teach the strategic thinking and research skills that will matter most in an AI-assisted workflow.

Third, what happens to the designer-developer boundary. I work at the intersection of design and development, and I see these worlds merging faster every year. AI tools that generate both design and code are accelerating that trend. The designers and developers who can work across that boundary will have a significant advantage.

The complement, not the replacement

The conclusion from my 2023 thesis was that AI tools should complement human creativity, not replace it. I wrote those words in an academic context, based on literature analysis. Now I’ve lived it for three years in practice.

The tools are better than I imagined they would be. The core insight is exactly as true as it was. Use AI to move faster through the parts of design that are about execution. Invest your human energy in the parts that are about understanding, judgment, and intent.

That’s where the real design work happens. And that’s where it will keep happening, no matter how good the tools get.

This is the final post in a three-part series. Read part one on what changed since 2023 and part two on evaluating AI design tools.