CEOInterviews.AI
Start App
Marissa Mayer
Former CEO of Yahoo, Lumi Labs

Marissa Mayer Wants AI to Read Your Camera Roll

📅 Sep 30, 2026 The Information 15 MIN 1351 VIEWS 29 SEGMENTS · 2 SPEAKERS
The Information Editor-in-Chief Jessica Lessin talks with former Yahoo CEO and Dazzle AI co-founder Marissa Mayer about her bet that your camera roll can make AI assistants more personal. Mayer explains how photos can give AI context about the people, places and preferences in your life—and how Dazzle is approaching the privacy tradeoffs. Subscribe: https://www.theinformation.com/subscr... The Information’s TITV airs weekdays on YouTube, X and LinkedIn at 10AM PT / 1PM ET. Or check us out wherever you get your podcasts. Follow us: X: https://x.com/theinformation IG:   / theinformation...

What Marissa Mayer said

Written from the verified transcript and checked against it. Every figure links to the moment it was said.

Marissa Mayer, former Yahoo CEO and founder of Lumi Labs, is launching a new AI startup called Dazzle, a personal assistant that learns from a user's camera roll. Mayer said Dazzle distinguishes itself by building a context layer on the user from their photos, enabling proactive actions like snapping a picture of a broken garage door and getting a repair service. She highlighted Dazzle's human-in-the-loop approach and separate credentials for agents to avoid sharing users' credit card or personal accounts. On privacy, Mayer said users can grant full, select, or no access to photos, and Dazzle deletes sensitive or personally identifiable photos it encounters. Mayer predicted the agent market will not be winner-takes-all, expects multiple agents, and discussed Dazzle's model-agnostic design, experimenting with newer model types. On monetization, she said subscription and transaction fees are likely. Mayer shared examples of Dazzle users, including a power user who got roofing estimates through the service.

Key takeaways

  1. Dazzle is a personal assistant that builds user context from their camera roll to enable proactive and personalized actions
  2. Mayer expects the AI agent market to have multiple winners, not a single winner-takes-all
  3. Dazzle gives agents their own accounts and credentials instead of transacting on user credit cards
  4. Dazzle deletes sensitive or identifiable photos it encounters, even with full access

Numbers and commitments

FigureWhat it refers toTypeAt
50,000 Mayer's own camera roll photo count metric 4:58
25,000 Average user's camera roll photo count metric 4:58

Chapters

  1. 0:00Dazzle's core concept
  2. 1:10How Dazzle works with photos
  3. 2:41Photo shareability insight
  4. 3:08Agents and human oversight
  5. 4:58Privacy and sensitive photo handling
  6. 6:14Lessons from Sunshine
  7. 8:23Agent market competition
  8. 10:22Model agnosticism and experimentation
  9. 11:25Business model and monetization
  10. 13:03Partnership approach and user examples

Questions asked in this interview

8
  1. 1:07So, how do you do it?
  2. 3:04What rash, what cream on this rash, if we're being super honest about it?
  3. 4:45How do you actually, how do you even begin to think about it?
  4. 6:01What did you learn, I'm curious, from Sunshine and sort of your first foray into this intersection of personal apps and AI and yeah, what worked, what didn't, and what are you bringing into Dazzle?
  5. 7:52... you think in addition to using the photo corpus, are there other ways that you think you guys will continue to stand out or am I thinking about this differently and will have multiple personal agents for multiple things two years from now?
  6. 9:47... I assume even in processing photos too, are we still in the world of LLMs or are you guys employing and have a view maybe on how some of the underlying capabilities and technologies in AI are evolving, are we still firmly in the LLM world?
  7. 11:16I can think advertising, I can think subscription. What are you thinking?
  8. 12:49Is because you're using humans, does that get you, do you need to have partnerships with all the service providers you're suggesting or routing to people or are the humans the interface there?
Interviewer 0:01 ↗
Thank you, Akash. And I'm so excited to be here with Marissa Mayer, who has a new startup, a new AI startup, that we'll talk a bit about, as well as just this crazy AI moment, Marissa, which I know you've been watching from so many angles. So, tell us about Dazzle and why we all need to make it another app on our phone.
Marissa Mayer 0:25 ↗
So, Dazzle is a personal assistant. And it's a personal assistant that learns from your camera roll. It does a couple of things that are different. There's obviously a lot of AI assistants right now. It's really having a moment. But we're different in that we build context on you and therefore can build a more personal AI because we understand who's important to you, what you like to do, where you like to go, how you like to spend your time from the history of your photos. And we also allow you to be proactive. So, if you need to get something done, you want to buy something, fix something, schedule something, you can just snap a picture of it, and we figure out what to do with it and bring you that action, or do it on your behalf with your permission.
Interviewer 1:07 ↗
So, how do you do it? Take us behind the hood a little bit.
Marissa Mayer 1:10 ↗
Sure. So, we have you install an app. You can also use Dazzle without the app. You can just text with us. You can use it on the web, but it works better with the app because what the app does is allows you to give us permission to look at your photos through your camera roll. So, you can give us full access. You can select the photos you like or no access at all. Obviously, it works best the more photos it has. And then what we do is we analyze your photos and we try and understand who you are. I've often said if I had to be your executive assistant for a week, I would ask for access to your email, which is what a lot of others are doing for context. We think Dazzle is the first in the space to say, 'Wait, if you really want to be a personal assistant, right, what do you like to buy? What should I book? Where would you like to travel? Right, what kinds of gifts do you like to give?' Having a sense of that person and what they think looks beautiful or what they want to capture memories of is really helpful. So that builds the context layer for us, which when we then make queries on your behalf either proactively or because you chat with us, we take that context layer and enhance your results. So for example, when I talk to Dazzle, it knows my kids' names, it knows my mom's name, it knows where I live, it does all that. And I never told it that. It just figured that out from my photos. And then the other thing it does is when you take a photo, we realized in the past I had worked on different applications that looked at photo sharing. But the average camera roll...
Interviewer 2:39 ↗
Google Photos among them, right? I think you were there.
Marissa Mayer 2:41 ↗
Right. And that one. Google Photos, Flickr, you know, Google Images. And you know, anyway, with all of those pieces, what we really found was that people's rolls are in large part not that shareable.
Interviewer 3:00 ↗
Mhm. There's a lot of things that you just want to remember.
Marissa Mayer 3:02 ↗
There's a lot of like...
Interviewer 3:04 ↗
What rash, what cream on this rash, if we're being super honest about it?
Marissa Mayer 3:08 ↗
I fix this? How do I address this? And we realized with that insight that there's a lot in the camera roll where without saying anything, right? The picture's worth a thousand words, we could actually take action for you. And so that's really the photo-based differentiators. I will say there's a couple of other things we do differently. One is that we have two ones in the loop. So we have our agents, they can do a lot for you, but when things get complicated or the agents get stuck or is something just more efficient to work with people, we have customer experience specialists that will step in and make sure that we can get that complicated task all the way to completion for our users. And the other is that when you're buying things and booking things, there's times where you need a payment credit card or you need to be able to sign in. And we think the moment is coming where you should probably be willing to give your credentials to AI, but we're not sure that's the moment quite yet. So, what we do is we give our agents, our AI agents as well as our customer experience specialists, we give them their own accounts. So, for example, they can buy you the tickets, and they can transfer the tickets to you, but it's not them transacting on your credit card. And so, between humans in the loop and having our own separate set of credentials for our agent, means when you go to buy something on Dazzle or book something on Dazzle, we come back to you, we get your confirmation before we actually transact, and then we transact in a way that you don't have to share your credit card with us or, you know, in any way, you know, RMS to impersonate you under the hat. We think that that moment is probably coming, but it's not here right now.
Interviewer 4:45 ↗
How are you thinking about privacy? Obviously, photos can be intimate. How do you actually, how do you even begin to think about it? Because so much of the utility is sort of knowing the full kit and caboodle.
Marissa Mayer 4:58 ↗
And the other thing is people's camera rolls are often really vast. Especially, you know, I don't, I mean, I have over 50,000 photos. We find the average person probably has 25,000. We even had one user that had more than a million photos on their camera roll. So, there's a lot on there. But what we've taken is, and our approach has been, one, you can decide what you want to share with us, full, select, or no access through iOS. But then the other thing is, we don't want your sensitive photos. You don't want us to have them. They're private and personal to you. But the thing is, with that many photos, we don't need every single photo to do a great job. So, our view is, when in doubt, leave it out. So, if while we're processing the photos, it seems like it's an intimate photo, it seems like it's got personally identifiable information, or just something else you wouldn't want us to have, we just drop it and immediately delete it. And so that's, you know, for the people who are doing full access, we still aren't accessing their sensitive photos. We're just discarding those as we find them. And we can still do a really good job. Done.
Interviewer 6:01 ↗
What did you learn, I'm curious, from Sunshine and sort of your first foray into this intersection of personal apps and AI and yeah, what worked, what didn't, and what are you bringing into Dazzle?
Marissa Mayer 6:14 ↗
Yeah, well, I think my excitement for this personal assistant moment goes back like two or three decades. So, I remember hearing about the idea of personal assistants back in grad school and even college and thinking about it would be so terrific. And a lot of different pieces have had to come together to really make this the moment and super exciting. I think that when I look back, you know, it's been wonderful to be an entrepreneur. So, you learn so much from the entrepreneurial journey about building companies and growing them. And I also think that there were some key insights from Sunshine. Namely, when we were working on Sunshine, one of the things we worked on was photo sharing. And that's when we were starting to look at our share-worthy model, our AI share-worthy model to try and understand how likely a person was to want to share a photo on their camera roll. That's really when we got the insight that that really wasn't where people needed the help. What they need the help on is taking those photos and transiting them into actions and doing it easily. Because the other thing is I don't know about you, but there's a lot of photos I snap and I forgot about. It's my son's birthday tomorrow. You know, about a month ago, he took me to his favorite football card store and said, 'These are the football cards I wanted.' I snapped pictures of them. And it's really great to actually have your assistant be like, 'You know what? You already know the perfect gift for your son. Like, here are pictures that you took.' And makes it really easy to do those types of things. So, having that proactivity that comes from being able to snap pictures and turn them into actions as opposed to going to a blank prompt box and saying, 'You know, what do I want to accomplish today' or sending it a text. Actually having that starting point of the photo, I find so helpful and it really reminds me of a lot of things I may have otherwise forgot.

15 more exchanges in this transcript

Sign in free to read the rest of this interview. No card required.

Sign in to read the full transcript

Cite this transcript

APA, MLA, BibTeX
APA

Mayer, M. (2026, September 30). Marissa Mayer Wants AI to Read Your Camera Roll [Interview transcript]. The Information. CEOInterviews.AI. https://ceointerviews.ai/interview/2985776/

MLA

Marissa Mayer. "Marissa Mayer Wants AI to Read Your Camera Roll." The Information, 30 Sep. 2026. Transcript, CEOInterviews.AI, https://ceointerviews.ai/interview/2985776/.

BibTeX
@misc{mayer2026_2985776,
  author       = {Marissa Mayer},
  title        = {Marissa Mayer Wants AI to Read Your Camera Roll},
  howpublished = {Interview transcript, The Information. CEOInterviews.AI},
  year         = {2026},
  month        = {sep},
  url          = {https://ceointerviews.ai/interview/2985776/},
  note         = {Speaker-attributed transcript with timestamps}
}