AI UX tools can make designers worse if they’re using them the wrong way

Why Read This from an HCI Perspective

This article doesn’t treat AI as just a ‘convenient automation tool’; it examines how AI changes UX decision-making, which makes it especially relevant for HCI practitioners and researchers. It also shows that the faster things get, the more important mid-course checks and accountability for explanations become. In research, it can lead to questions not about the accuracy of AI outputs, but about when people choose to trust them, when they doubt them, and how they decide to intervene.

CIT's Commentary

The most interesting point is that AI doesn’t degrade quality by ‘making mistakes’ so much as by taking away judgment. In UX, the hard part is rarely ‘finding the right answer’; it’s enduring ambiguity and redefining the problem. If AI removes that discomfort too quickly, teams may feel comfortable settling for average results. So the key isn’t the tool’s raw performance—it’s designing for intervention paths. In particular, from a research perspective, even if you use LLMs to build UX research summarization or evaluation tools, you need accompanying mechanisms that reveal what the tool is missing. What matters more than fast generation is creating a structure that makes people ask questions again.

Questions to Consider While Reading

• Q.What kinds of interface mechanisms would help people see AI-generated UX results not as ‘answers,’ but as ‘questions’?
• Q.When building research summaries or UX measurement tools with LLMs, how should you balance accuracy with interpretability?
• Q.In product organizations where speed is critical, how can you institutionalize the points where humans must intervene through policies and workflows?