AI has made product work feel weirdly cheap in the middle and expensive at the edges. Another screen, prototype, or feature is easier to produce than it used to be. The harder work is deciding what belongs, what can be trusted, and what should never ship.
That is a different kind of leverage than pushing pixels faster.
Karolina Rojek, writing in UX Collective, names the design problem underneath the AI boom:
As technology accelerates, the limiting factor is no longer our ability to build. Increasingly, technology can help create itself. The harder question is deciding what should be built and how it should fit into people’s lives. That is a human problem, and design sits at the center of answering it-Jon Friedman said.
Design is now the discipline of deciding what should be made, who it is for, what behaviour it encourages, what it removes, what it simplifies, and what kind of relationship it creates between people and technology.
Rojek’s piece tracks Microsoft, Samsung, Shopify, Meta, OpenAI, and even the federal government elevating design leadership. That list is easy to read as corporate signaling, but the pattern is more interesting than that. If AI makes output cheaper, design has to move upstream: what should this product do, how should it explain itself, when should it stay quiet, and when should it hand control back to the person using it?
Rojek:
One of the traps product teams fall into is thinking the interface is the product.
The interface is where the user touches the product, but the experience is shaped by everything around it: onboarding, pricing, permissions, support, notifications, failure states, data policies, service handoffs, organisational incentives, and the user’s real-life context.
AI makes this even more obvious.
She puts that into everyday product terms:
Imagine a small business owner using an AI tool inside an e-commerce platform. The value is not simply whether the AI can generate a product description. The value depends on whether the merchant trusts the suggestion, understands how it affects SEO, can edit it easily, knows whether it matches their brand voice, and feels confident publishing it.
Or think about a designer using an AI prototyping tool. The value is not only speed. It is whether the tool helps them think more clearly, explore better alternatives, communicate intent, and avoid creating generic work at scale.
Or consider a citizen trying to renew a passport or access benefits through a government website. The problem is not just visual design. It is language, accessibility, anxiety, eligibility rules, documentation, error recovery, and the emotional weight of dealing with a system that may affect their life.
This is why “powered by AI” is such thin product strategy. The model is only one part of the experience. Trust, permissions, support, and failure recovery decide whether the product feels useful or feels like more work.
Rojek on craft:
AI can help generate options, but craft helps decide which option has integrity.
This is especially important because AI products can easily become noisy. More prompts. More suggestions. More assistants. More summaries. More generated content. More things asking for attention.
Without strong design judgement, AI can turn products into cluttered systems full of technically impressive but emotionally exhausting features.
Cheap output and coherence are different problems. A product can generate options all day. Someone still has to decide which options deserve to exist, which ones ask too much of the user, and which ones quietly turn into another obligation.
