AI in UX: Where to Start, and Where to Stop
As artificial intellegence takes on more roles, how do we decide what tasks still require a human touch?
LET’S DISCUSS YOUR NEXT PROJECT
AI is everywhere. It’s rewriting job descriptions, generating product copy, and showing up in almost every LinkedIn post I see.
But as UX folks, we have to ask a better question than “what can it do?”. What we all need to be asking is, “what should it do?”.
Yes, there are joys and drudgeries in every job. But that doesn’t mean it’s a good idea to farm out all the tedious work. There’s a future where we get close to that, but first we need to figure out what we want AI to take on - and what we need humans to hold on to.
Take notetaking: There’s hardly a meeting I attend that isn’t joined by an AI notetaker. I take notes religiously, sometimes on paper, sometimes digitally–never with AI... When I need to refer back to my notes, they can be invaluable; and AI has that covered for all of us. But the real value of notetaking, in my opinion, lies in the mental processes triggered by the practice–the act of taking notes encodes the information into memory and triggers thoughts and insights that may otherwise lie dormant.
As of today, here’s where I draw the line.
Image of golden robot with bright blue eyes inquisitively putting a mechanical finger to its brow. [Photo by Zia Halawa]
Where to Start: Where AI Might Help Today.
There are plenty of areas where AI has real potential to make UX better; here’s where you should start exploring.
Information triage in data-dense environments
Let AI surface anomalies. Flag things worth a second look. Highlight data that’s changed meaningfully. That doesn’t mean it decides what’s important on it’s own, but it can raise the cases we need to manually review.
Accessibility testing and simulation
Use AI to scale accessibility reviews. Run models to detect contrast issues, screen reader gaps, and tab order failures. Then let the real humans using assistive tech as the final gold standard of testing.
Smart defaults and form field guesses
AI can be useful for reducing friction, like pre-filling a form field or suggesting a configuration. Just make sure it can be changed and isn’t locking a user in to anything.
In these areas, AI is like a new hire - shows promise but may not be ready to own the process completely. Use it, but keep an eye on how it's progressing over time.
Where to Stop: Where AI Has No Business Being
And then there are the places where AI doesn’t belong - and we don’t see that changing.
Usability testing
Simulated users don’t rage-click. They don’t hesitate because something “feels weird.” They don’t bring the baggage of a long workday or the stress of a tight deadline. Real UX feedback comes from real people in real contexts. No shortcut replaces that.
Design strategy and product vision
AI can help you analyze what’s happening. But it doesn’t know your business goals, your market pressures, or what your users actually care about. Design direction - deciding what’s essential, what’s fluff, and what’s next - requires human judgment and a bit of gut instinct that’s not in an AI tool.
User judgment
In high-stakes environments - think healthcare, logistics, finance - automating user decisions is too risky. Good UX gives users the information to make better decisions, it doesn’t take the wheel.
What AI Should Do
AI should handle the repetitive stuff - sorting, flagging, filling, filtering. Tasks that are time-consuming, repetitive, rules-based, and don’t require human judgment.
It should highlight patterns, not make assumptions. Suggest next steps, not take them.
It should improve workflows, not hide them behind a black box.
If you're adding AI to your product, start with this filter: Does AI make the product better, or simply get you to the finish line faster?
The goal isn’t to skip the design process, it’s to sharpen it. AI should amplify a designer’s expertise, not stand in for it.
LET’S DISCUSS YOUR NEXT PROJECT
We look forward to hearing from you.
