“We built a process called silent network authentication, where we work with mobile carriers to authenticate a user seamlessly, with no SMS code. Technically it's far more secure and frictionless. But in early user studies, because it was so silent, a lot of users thought it was less secure. They missed the friction. We had to design a UI element that says, 'We are securely authenticating you in the background right now.' Transparency builds trust.”
Insights Unlocked · User Experience · November 2026
Context
Asked where AI delegation goes from helpful to unsettling, Dodeja says the line is human psychology, and uses a user-research finding to show that invisible security can feel less trustworthy.
Transcript
…you don't want is somebody pretending to be you and sells your tickets to somebody else or steals your tickets to somebody else. So the trust becomes really critical. And having a very clear authorization layer that both the businesses can trust and the human can trust is actually very critical.
…behaviors and interaction models that are still unfamiliar to a lot of people, how can product and UX teams understand where customers are comfortable delegating decisions to AI and where they want to remain in control or where an experience crosses the line from being helpful to being unsettling?
Yeah, I think that's a core UX challenge right now for our generation, right? As we transition from tools we use ourselves and having AI and delegate that to AI agents, right? Software moves from executing demands that we were making on our behalf versus somebody else doing it on our behalf. And I think there's a fine line between magical and creepy, being creepy there, right? So I think the way I think about it is that products and UX teams have to figure out where their users are comfortable delegating versus where they want control. So I think of it like they need three things, right? So one is figure out what is the low stakes delegation stuff, right? Users are very comfortable delegating tasks where the cost of a mistake is near zero. The cognitive load is super high. Think of rescheduling a meeting, finding flight options, organizing an email. In this case, users want the AI to execute and give them a quick summary after the fact. In fact, personal story I'll say is: I need to find new home insurance in the state of California recently. So what did I do? I went to my favorite AI agent and asked it to find me options. And it made a whole bunch of some, made a few calls, opened a few websites, stuff like that. But I was able…
Yeah. And to that point, and speaking of About how irrational we all are. Like, we just have to kind of, and all this is such new territory, right? That we have to be very cautious now. But five years from now, we will not be, right? We won't need to have all those sort of things. And you have to…
Yeah, like we don't think about dial-up internet anymore.
Right. What else should product leaders be thinking about as they think about security and identity as they're building products? And what should they keep going to the customers to try and understand as they do this? Yeah.
More from Insights Unlocked
