The future of AI in UX
A lot of the AI conversation in design still feels off to me.
It’s either “this changes everything” or “this is all overhyped,” and neither is that helpful when you’re actually building products. The reality is probably more in the middle. AI is already showing up in real ways, but most of it is still rough. Some of it genuinely improves the experience, but some of it just adds noise.
That’s been my takeaway thinking about this over the past several years, especially after running an AI-focused design-athon at Grubhub. We came up with a lot of ideas that sounded smart in a room, but few of them held up once we started asking, “would someone actually want this?” or “does this make the product feel better or just more complicated?”
That gap is where most of the work is.
Take personalization for example. We’ve been talking about it forever, but most personalized experiences are still pretty blunt. It’s usually just “you ordered this before” or “other people like this.” I think AI can make that better, but only if it’s applied with some restraints.
At Grubhub, we explored this with menu recommendations. On paper, it’s easy to say - use order history, time of day, weather, and just surface better options. But then you get into it and realize how quickly things can feel repetitive or even a little pushy. If I ordered the same thing three times, do I want to keep seeing it? Maybe. Maybe not! If the system gets too confident, it actually narrows the experience instead of improving it.
I think that’s the part that doesn’t get talked about enough. Just because a model can make a good prediction doesn’t mean it makes for a better product.
The same thing is happening with AI in the design process, too. There’s a lot of excitement around speeding things up, generating layouts, running tests, all of that. Some of it is genuinely useful and there are definitely moments where AI helps you get unstuck or explore faster. But, it also makes it really easy to produce a lot of average work. Faster doesn’t automatically mean better. If anything, it puts more pressure on having a clear point of view, otherwise you just end up with more options and no real direction.
For me, the value of AI is pretty simple. If AI can take some of the repetitive work off the table, good! That’s time back for strategic thinking, shaping the product, or figuring out what actually matters to users. That part of design doesn’t go away…
The part that still makes me uneasy is the trust. It’s easy to build something that feels helpful on the surface but is doing things the user doesn’t fully understand? Think about recommendations, ranking, even small UI changes driven by AI - they all carry assumptions. If those assumptions are wrong, or biased, or just unclear, then users feel it pretty quickly.
You see it when something feels off. Like when a recommendation is technically relevant but just weirdly timed, or when a system keeps pushing you in a direction you didn’t choose. It can be subtle, but it chips away at confidence in the overall product or brand.
I don’t think designers can act like that is someone else’s problem. If anything, this creates more design work to be done, not less. We should be thinking through how these systems behave, when they step in and when they should back off, or how much control the user actually has.
Not everything needs to be automated and not every decision needs to be predicted.
We all know AI is going to keep showing up in products, that part is pretty clear. The hard part is being selective about where it actually helps us. Today, the best experiences I’ve seen don’t draw attention to AI at all. They all just feel a little easier, a little more relevant, a little less effortless…
and that’s probably the new bar.
