Stefanie Hutka

Things Stefanie Says on Podcasts

Design researcher with a PhD in cognitive neuroscience who works on how teams decide what to automate. Runs the consultancy Sendfull, teaches UX design and systems thinking at UC Berkeley, and previously led UX research for product launches at Adobe, Meta Reality Labs and DAQRI. Writes the biweekly Sendfull newsletter on Substack and is the author of What Your Machines Should Do (Rosenfeld Media). Guest on Finding Our Way and the Rosenfeld Review Podcast.

Where to Find Them

Stefanie Hutka has been a guest on Rosenfeld Review Podcast , Product Impact Podcast | Secrets to unlocking the value of AI and Finding Our Way .

Recently: “AI will Automate All Tasks. Don't Let it Automate Your Job. (Stefanie Hutka, PhD)” on Product Impact Podcast | Secrets to unlocking the value of AI (November 2026); “What Should Your Machines Do? with Stefanie Hutka” on Rosenfeld Review Podcast (September 2026); “77: Automate All the Things!? A UX Reality Check on AI (ft. Dr Stefanie Hutka)” on Finding Our Way (September 2026).

What They Said

A host asked what people overestimate that LLMs are good at. Hutka calls automation bias one of her favorite human-factors findings and warns that AI outputs look convincing unless a system is designed for verification.

Product Impact Podcast | Secrets to unlocking the value of AI · 2026-11-11 Permalink → Listen →
Product Impact Podcast | Secrets to unlocking the value of AI Around 43:11 into the episode
Stefanie Hutka

…to kind of extend your reach, or if you're working in a bigger company where you're working with, you know, cross-functional partners and machine learning engineers, that will help you be more effective in terms of communicating your domain expertise and kind of translating across disciplines.

Brittany Hobbs

And again, in terms of the hot takes here, what is something that you think people underestimate in terms of their own capability and agency that is far superior to LMs? And what do you think is something that people overestimate that LLMs are good at and actually is terrible at?

Stefanie Hutka

Overestimating a lot. So there's this thing called, this is probably one of my favorite, like, I don't know, fun facts from human factors, like automation bias is real. We tend to believe the output of an automated system more than our own judgment or that of others. And this, you don't even have this with, you know, it's like not, you don't need 10 agents running and giving you output for this. Like if you've ever followed Google Maps, just, you know, hey, there's a lake in front of me, but it says to keep going. Like that is a good example of automation bias in practice. So I think that is just something. need to be aware of kind of at all times using…

Arpy Dragffy Guerrero

you mentioned about notebook lm or no gemini notebook also great because you can also put in multiple language assets and then query it in english fabulous what other products do you genuinely love using like ones that leave you feeling more capable rather than more managed or needing to manage

Stefanie Hutka

yes i'm gonna throw out something that is maybe a bit orthogonal and it's gonna be my e-bike and there's two reasons i mentioned this uh so you know i i commute from oakland to to berkeley twice a week and i can sort of choose my own adventure there's still autonomy if i want full pedal assist and…

Arpy Dragffy Guerrero

thank you stuff so much for joining us so you have a book your book is what your machine should do coming out on november 10th so first and foremost where should everybody buy that

Speaker names come from our own diarization. The position is estimated from where the line sits in the episode.

Collections They Appear In