That distinction matters more than it sounds like it should. AI is becoming advanced enough that people are beginning to treat it like it has judgment. It doesn't. It's good at recognizing patterns, but recognizing patterns isn't the same as understanding context, business goals, or the people you're designing for.
AI hallucinations are the clearest proof of that gap. In healthcare transcription, for example, AI-generated errors have sounded plausible enough that nobody thought to question them, even though they were simply wrong. Research on these tools backs this up: many would rather generate a confident answer than admit they don't know. The mistakes aren't the surprising part. We all know AI gets things wrong. What's surprising is how believable the mistakes sound. Unless someone with experience is checking the output, a confident error can quietly become the insight everyone starts building from.
The interesting question isn't what AI can do. It's what a good designer can do more efficiently, and even better, because they know exactly what to ask AI for and exactly what to throw out.
A few patterns worth knowing:
AI can turn unstructured, qualitative data into organized themes fast. It codes interview quotes and maps them for actual conceptual closeness rather than just similar wording. But AI only organizes the data. A researcher still has to build the narrative: providing the context, connecting the dots, and deciding what actually counts as a finding.
AI can help a single designer scale an existing design system quickly, running something close to "design ops" with far less manual work. But this only works because the underlying component library is already thorough, the flows are well-defined, and the page layouts are established. Hand AI a blank page and it has no idea what "on brand" means.
AI can help build out mood boards fast enough to give a team something to react to, instead of starting from a blank page. But someone still has to pull the initial concepts and evaluate the results against heuristics, design trends, and brand context. AI can produce options. It can't judge which ones actually fit.
Across all three, the same idea holds: AI does the typing. The designer does the thinking.
We've seen the same thing in our own work. AI has become a valuable part of our research process because it helps us organize interviews and user testing results, validate emerging themes, and pull supporting quotes from transcripts much faster than we could manually. What used to take weeks now takes days to synthesize.
That speed only works because we've already established solid research framing and methodology. AI helps generate a strong first draft of the analysis. That's exactly what it is: a first draft. Before anything becomes an insight or recommendation, we go back to the interviews through the lens of business context and human understanding, and compare every theme against what customers actually said.
If anything, AI has made research judgment more valuable, not less. AI can surface themes across dozens of interviews in minutes, but it also overemphasizes ideas that weren't actually representative of what we heard in the research. That instinct sharpens with practice. It never runs on autopilot. What improves over time is your sense for which findings are worth a second look.
Before handing a transcript, feature audit, or early product flows over to AI, ask two questions:
If you can't, that's exactly where a confident but wrong answer has room to slip in.
AI is changing how UX work gets done. It isn't changing what solid UX requires. The technology is evolving quickly, but human judgment is still what separates useful insights from convincing guesses. That's the discipline we help clients build into their own process, before a confident AI answer becomes an expensive one. If you're scaling AI into your team's research or design work and want a second set of eyes on where the real judgment calls sit, that's a conversation worth having before the rollout, not after.
That distinction matters more than it sounds like it should. AI is becoming advanced enough that people are beginning to treat it like it has judgment. It doesn't. It's good at recognizing patterns, but recognizing patterns isn't the same as understanding context, business goals, or the people you're designing for.
AI hallucinations are the clearest proof of that gap. In healthcare transcription, for example, AI-generated errors have sounded plausible enough that nobody thought to question them, even though they were simply wrong. Research on these tools backs this up: many would rather generate a confident answer than admit they don't know. The mistakes aren't the surprising part. We all know AI gets things wrong. What's surprising is how believable the mistakes sound. Unless someone with experience is checking the output, a confident error can quietly become the insight everyone starts building from.
The interesting question isn't what AI can do. It's what a good designer can do more efficiently, and even better, because they know exactly what to ask AI for and exactly what to throw out.
A few patterns worth knowing:
AI can turn unstructured, qualitative data into organized themes fast. It codes interview quotes and maps them for actual conceptual closeness rather than just similar wording. But AI only organizes the data. A researcher still has to build the narrative: providing the context, connecting the dots, and deciding what actually counts as a finding.
AI can help a single designer scale an existing design system quickly, running something close to "design ops" with far less manual work. But this only works because the underlying component library is already thorough, the flows are well-defined, and the page layouts are established. Hand AI a blank page and it has no idea what "on brand" means.
AI can help build out mood boards fast enough to give a team something to react to, instead of starting from a blank page. But someone still has to pull the initial concepts and evaluate the results against heuristics, design trends, and brand context. AI can produce options. It can't judge which ones actually fit.
Across all three, the same idea holds: AI does the typing. The designer does the thinking.
We've seen the same thing in our own work. AI has become a valuable part of our research process because it helps us organize interviews and user testing results, validate emerging themes, and pull supporting quotes from transcripts much faster than we could manually. What used to take weeks now takes days to synthesize.
That speed only works because we've already established solid research framing and methodology. AI helps generate a strong first draft of the analysis. That's exactly what it is: a first draft. Before anything becomes an insight or recommendation, we go back to the interviews through the lens of business context and human understanding, and compare every theme against what customers actually said.
If anything, AI has made research judgment more valuable, not less. AI can surface themes across dozens of interviews in minutes, but it also overemphasizes ideas that weren't actually representative of what we heard in the research. That instinct sharpens with practice. It never runs on autopilot. What improves over time is your sense for which findings are worth a second look.
Before handing a transcript, feature audit, or early product flows over to AI, ask two questions:
If you can't, that's exactly where a confident but wrong answer has room to slip in.
AI is changing how UX work gets done. It isn't changing what solid UX requires. The technology is evolving quickly, but human judgment is still what separates useful insights from convincing guesses. That's the discipline we help clients build into their own process, before a confident AI answer becomes an expensive one. If you're scaling AI into your team's research or design work and want a second set of eyes on where the real judgment calls sit, that's a conversation worth having before the rollout, not after.
