#58 - Android's big leap is from apps to agents - Sameer Samat

The smartphone has been built around apps and taps for nearly two decades. Google thinks AI will fundamentally change that.

In this episode of The Deep View Conversations, we talked with Sameer Samat, president of Android ecosystem at Google, about what the company means when it says it's transforming Android from an operating system into an intelligence system. 

Samat explains why the next generation of computing could shift us from micromanaging our devices to simply telling them what we want to accomplish. We dig into how AI agents could navigate apps and complete multistep tasks and why those agents need to follow us across phones, computers, cars, watches and glasses. And what happens to the app-centric model that has defined smartphones for the past 15 years?

We also get into some of the practical ways this is already taking shape. Samat discusses Google’s app automations and Rambler, the new Google Keyboard experience that can turn your voice brain-dumps into polished text. He also explains how Google is thinking about permissions, sandboxing and human oversight as AI agents gain the ability to take action on our behalf.

The conversation goes well beyond the phone. We talk about why smart glasses and cars could be especially powerful interfaces for AI agents, what Google learned from the original Google Glass, and why the best AI features may be the ones consumers barely think of as AI.

Other topics covered include:
• How AI is already changing work inside Google
• Why product managers can now build functional prototypes themselves
• Samat's favorite overlooked AI tool
• His "calendar cleanse" strategy for getting time back

If you’re trying to understand where mobile computing goes next, what AI agents will actually look like on phones, and how Google plans to weave intelligence across devices, this conversation offers insights into what the company is building and why.

Subscribe to Deep View Conversations for interviews with the leaders shaping the future of AI, business, and technology: tdv.transitor.fm 

And don't forget to sign up for The Deep View daily newsletter. We don’t just cover AI, we decode it. In a world flooded with hype, we deliver sharp, no-nonsense insights to keep you ahead of the curve and help you put AI to work every day: subscribe.thedeepview.com

Creators and Guests

Host

Jason Hiner

Editor-in-Chief of The Deep View

What is The Deep View: Conversations?

From frontier labs and enterprise platforms to emerging startups reshaping entire industries, The Deep View: Conversations podcast interviews the brightest minds and the most influential leaders in AI.

Jason Hiner: Well, Sameer, it's good to see you.

Sameer Samat: Nice to see you too. Yeah.

Jason Hiner: Thanks for taking the time.

Sameer Samat: Of course.

Jason Hiner: Big day, Made by Google. This was one of the big events of the year for you all. There's a lot happening with Gemini intelligence, with the ways that the device itself, these devices themselves, are doing a lot, a lot more, and a lot kind of differently too. So, you've talked about the way that the device is moving, Android is moving from an operating system to an intelligent system. What does an intelligent system do that an operating system doesn't do?

Sameer Samat: Yeah. Well, of course, thanks for having me. I really appreciate the conversation.

Jason Hiner: Yeah.

Sameer Samat: I think that from the standpoint of an operating system to an intelligent system, maybe the place to start on that is just the way that I view what we've been trying to do with computing for the last 40 years or so. That's really try to make computing as easy for people to access and utilize, but in the spirit of have all of us learn how best the computer takes instructions, and that paradigm has just played out over a number of different interfaces. So, you have command line interface, which is, there's a set of commands you can issue and what have you. I mean, you start with assembly language and you command line interface, and then eventually you have graphical user interfaces, which are helpful because they broaden access and it makes it easier. But you're still taking a pointer and you're clicking on individual actions you want to take and you have menus and all these things that are supposed to be organized, but they're the computer understands. I think all of that needed to be managed by a system and from a resource standpoint, what have you, and we call that the operating system. If you think about where computing is going, it's really more flipping the paradigm to actually the machine can understand your goals, your intention, and it can start to transition from taking micromanagement from you to instead you are articulating like, I want to get this done. And then it translate that into a whole bunch of actions, right? And I think that's fundamentally a different human-computer interface. It's very exciting. And I think that vision is what we call an intelligent system. And so of course, the same if you have the operating systems book from your computer science class still on your shelf, we still have to do all those things that you have managing resources and all those things. But on top of that, there needs to be this intelligent layer, which is able to take what your goal is and translate that into all the things that you previously had to do yourself. And that's our vision. And we are starting out with that with Gemini intelligence on Android and moving in that direction. We're launching. So we've launched a few things around that already. We have more coming this week.

Jason Hiner: Yeah, very good. When we talk specifically about mobile, the center of the mobile ecosystem has been the app for the past, you know, 15 years, 20 years. In an intelligent system, the app seems to be a little less centered and the things like the ways that you all now make app functions accessible in a more kind of atomic way. So atomizing some of the functions of an app in making them available to intelligence. That's a different way of approaching it. How does that relate to kind of the way you're thinking about it and the way you're talking about, you know, the way we access these systems?

Sameer Samat: Yeah, that's a great question. I'd answer it with two different things. First, I think the app is at a bit of like a meta level. The app is a set of functionality packaged by a particular provider of those services. Right. And so I think fundamentally we say, OK, you know, is the app going away? If we mean sets of services from folks that have a brand that you interface with, I don't think that's going away. I don't think it went away when we moved from mobile to apps. I don't think it's going away now. Yeah, the form of it could change in some regards. But the second part I wanted to say is we should define what from it, from the standpoint of what people are actually doing on their phone. So if you look at time spent, you know, on a phone, a lot of it is spent doing things that I don't think we want agents to really replace. A lot of it is playing games. A lot of it is watching, you know, a movie in YouTube or looking at TikTok or Instagram. And that's that's over 70 percent of time spent on phones. So just for those sets of things, with all those things that are consumption oriented, playing games, interacting, messaging with your friends, you know, like those are that's the majority of the time we spent. So actually, when we when people talk about apps going away and all this, I like to just ground it in this notion of a two by two, if you will. And one dimension of it is experience. And the other dimension is how transactional the app is. Right. And if you think about the dimension of experience, like the higher the experience it is in the app and the lower the transactionality of the app is, the more it's probably not going to be affected by all this change, if you see what I mean, because we don't really want our I mean, maybe we should have our agents doing some of these shows instead. But like we kind of want to watch a lot of these things on YouTube. And, you know, we want to learn. I watched one of your your videos the other day, you know, I don't want an agent to do that for me. I want to do that. Right. That's why it's there. Right. And so I think that but if you get so if we want to talk about, you know, the the when you ask that question about apps, I think it's really referring to the things that are high on a transactions and then low on the experience scale. Right. Yep. And yeah, I think there's a possibility there that those that those companies will say, hey, you know, I want to I want to have more of my functionality be available to agents and to systems that can can help automate these things because it actually increases the pie of things that can be done. You know, if if I'm wearing my intelligent classes or I'm in a car and I want to order something, you know, reorder my regular from whatever, you know, that that's a transaction that I might be able to make now that previously I can't fumble around and get it done. And so I may not do it. So I think it's actually expansionary for those folks if it's done right. And for the rest, I think it's to be determined, but I don't think it's obvious that the things that we want to consume are going to be gentrified away.

Jason Hiner: OK, you know, that makes a lot of sense. You know, there's there's these interesting paradigms and I want to acknowledge, too, like it's different in different places, even regionally. So like in Asia and China, there's been more like the soup wrap that does a lot of things for you. And in more like US and Europe, it's been a lot of apps that do you know, we sort of have the app that we pay for things with or, you know, part of the OS. We have an app that helped us get rise. We have one that we do, you know, other things with. So the paradigms are different, but they seem to be shifting all, you know, all of them seem to be shifting as we think about and we sort of play forward some of these advances in the ways things have been evolving, you know, in the ecosystem. Today's the way people use AI and their phones is still mostly a chat bot. Right. What what Google is talking a lot about, and I expect others to talk more about, too, is this idea of agents on the phone. You know, you reference is a little bit how does how's that play now in this new generation of devices of the new generation of the OS that that you all announced at Google I.O. What's the vision of agents on phones, you know, in twenty twenty six and

Sameer Samat: twenty twenty. That's a great question. So I think with this comes back to how we view the intelligence system. And I think that, you know, what we would like to enable for people is a system where, you know, the chat bot interface, I think is a good one. We'll continue. People seem to vibe with it and enjoy it. But I think for long running tasks, for things that you actually may take some time to complete, you know, you need an ability for the system to help you understand, first of all, to be able to accomplish those things for you. You know, so, you know, if you want it to go do a series of actions for you that represent you know, a thing you're trying to get done, you know, like I need to figure out when soccer is and then I need to make make sure that we send the right messages to the right people because we have to do the carpool at this time. That's a series of steps. And so we've started with that already with some functionality called called app automation in Gemini with Gemini intelligence. And it is this makes Gemini capable of taking a task and then launching different apps and actually navigating them and performing those kinds of things for you. It's just the beginning of that. But we view it as pointing towards the direction of, you know, these systems can help you with those kinds of things that give you more time back in your life. And that's cool. And then there comes the next question of, well, what is the user interface for that look like? Because it isn't always, even if you had a personal assistant that you were hiring and paying by the hour, that person may have questions for you. Sure. You know, they may want to interact with you, you know, they may want to say like, well, you have two choices, A or B, you know, and did I get this right before I pay for this? Like, can you check this out?

Jason Hiner: You know, yeah.

Sameer Samat: And so how do you create an interface for that? And so those are all parts of like the intelligence system that we think about and then and the new ways you may interact with your phone. By the way, not just your phone, you will probably want to continue these interactions depend no matter what device you're on.

Jason Hiner: Cross devices.

Sameer Samat: Absolutely. And you want it to be the same personal assistant you're talking to, right? You don't want a personal assistant on your watch that's different from your phone because, you know, you want to continue those tasks. So that's how we think about it.

Jason Hiner: You can see yourself getting really frustrated if you're like trying to coordinate right on your phone, then you go sit down and computer and you're working and you're like, OK, now that thing that I told you earlier, can you change that to two o'clock and it doesn't know, you know, it would get frustrating.

Sameer Samat: I would get frustrating. I would get frustrating. And and I think the notion of a personal assistant would be one that has enough context about you with your permission to be able to help you with these tasks and then be in all the places that naturally make sense to be in. And so certainly your phone is great interface. Your laptop may be a great interface as well. You know, when you're on the go and driving somewhere, you know, you want to be able to ask your car and want to have that conversation continue. So that's really an important part of the continuity that makes it feel smart, intelligent.

Jason Hiner: Is that why is that where you went to Gemini intelligence? Because the idea that you have a Gemini app that is in all of those spaces, desktop, car, phone, glasses, watch, it can extend across all of those.

Sameer Samat: That's pieces that a Gemini intelligence for us is basically the best Gemini experience on the most advanced Android devices. And, you know, of course, the phone is an important centerpiece for all of us, but there's many other devices that, you know, that functionality should be extended to. And I think the agent and a personal assistant is one element of this. But, you know, you were asking, you know, hey, what is this intelligence system earlier? And I just wanted to give like a very practical example for your listeners of something we're launching this week that I think is a great example of going from this paradigm of micromanagement of the machine to one where you can be goal more goal oriented. And that's what we're doing with the keyboard. So we have this new microphone in the keyboard that we're launching. The internal name for has been Rambler. And it's pretty exciting to me because, you know, historically, the keyboard has been a an input system that was really in part designed for the machine. Right. And and it is I'm in an alphabetical order, you know. And and so from a human standpoint, you really have to learn to use this thing. People go to typing class on whatever. Right. And you're taking literally taking like your thoughts. And then you're like, OK, I'm going to break this down into what I want to say. And then I'm going to break the words down into letters. So this is inputting into the machine and learning how to do that, like to a T. Right. So what the so we we've of course had dictation microphones. That's not new. You press a button and you dictate, you know, but that is still pretty micromanagey because you're literally faithfully dictating every single word. Oh, yeah, I didn't mean that. Maybe this instead. Oh, wait. OK, so you're you're really having to be careful what you're dictating and go back and edit. Ramblers different. So this is open the microphone on your keyboard. Speak ramble, if you will, do a brain dump and it will come back to you and basically and and synthesize what you what you meant for for you to review. And I've been using it for two months. It's totally changed the way I use the phone. I don't speak multiple languages regularly, but for those who do, it is very common for people to speak out loud and go between like your charts, a different language. Yeah, it's very common in many parts of the world where, you know, you go back and forth, remember, can handle that. That's why it supports over 100 different languages, you know, and and it can do it all in one language for you or it can have both languages in there. So I think it's really a great example of the keyboard becoming intelligent. And and that's one component of overall redesigning operating system to be an intelligent system.

Jason Hiner: You're reading my mind like an agent now because I was going to go next. We talk. You brought this example of like, say I have like this afternoon. I want to go to see this afternoon. I have to do a pickup like a carpool pickup at one o'clock for soccer or something. And so I just want it. I want to. So there's a number of tasks for me to do. Look up the calendar and see the people were in it. Put it past text them, coordinate where to meet, things like that. So the dream with this is to be able to say to send to the agent, hey, send a calendar invite for one o'clock, you know, with the group of carpool, do the same one we did last week. The same group of people and all that. And what I was going to say is like. But you don't really want to type all of that in. It will be much nicer if you could just say it and do it in an unstructured way. Because because our thoughts are often unstructured, right? Like we don't have to think about it sequentially. And then have the A.I., which is A.I. is good at providing structure, unstructured things. Let the A.I. when that's what Ramblers doing, it's providing the structure and then it's going and carrying out the steps. Absolutely.

Sameer Samat: And I think that the the great thing about these multimodal models is, you know, we, we express ourselves in lots of different ways. Typing or text input is one way, but, you know, we like to just talk.

Jason Hiner: Yeah.

Sameer Samat: There's a reason why I think this notion of a brain dump has been a human thing for like a very long time, like I guess sometimes you just need to say to somebody else, like everything you're thinking and just the act of saying it helps, helps in some ways by the end, not during, but by the end, clarify what you really meant, because a lot of these systems are, you know, they're, they can goal seek, but what is your goal?

Jason Hiner: Exactly.

Sameer Samat: Like, what are you exactly saying you want to do? And I think we found this with coding too, you know, with these systems. Right. And so actually I've observed a lot of people who are using these systems for coding with a boom mic now, you know, and they're, and I think there is something to saying it out loud and what have you, but just dictating that may not at the end, you're like, okay, wait, this is like a mess up yet. It's like, Hey, take the gist of it. Try to understand, synthesize it, give it back to me. It's like, yeah, okay. That's what I meant. And so I think that's a big step. We're super excited about that launching on, on pixel 11. I think it's, uh, I can't wait for people's reaction to it. I think, um, you know, it's been a really anticipated feature. I get a lot of emails about it. When is this coming? So, um, yeah, we're excited to see that pull out.

Jason Hiner: Yeah. I've been one of those email and people Google about like, can I get a rambler? Can I get a look at it? Because I like, cause you can, um, if you don't know, if you haven't already, like you could sign up with the beta in Google play store, like you can go to the app and ask to be in the beta of, of Google keyboard, G board.

Sameer Samat: Yeah. So, so you need to get G board. Um, and once you get G board and you have pixel 11, it'll roll out. Um, one of the cool things, of course, because it's G board, right? So it's any input field, right? So it's not just like in the Google apps. Like anytime there's an input field you tap and the G board comes up, you could tap that mic and then you can just grapple into it.

Jason Hiner: A lot of people so far in 2026 have gotten excited about whisper flow. It has a very similar, um, uh, approach. Uh, so very much looking forward to that. A lot of people that I talked to the AI ecosystem also really, uh, excited for that because one of the challenges with a third party tool like whisper flow and there's others, you know, as well, um, is that sort of, it has this little bit of herky jerky experience of like, you sort of have, sometimes you have to like go out and then go back in, right? The nice thing about it being Rambler being in G board is it's always right there. It's similar to what would have been in the past, like the, the microphone button. Um, but now it can sort of take this unstructured, you know, stuff that comes out of here to there and, and make it a lot better.

Sameer Samat: You know, G board's been around for a while. We have so many people, uh, we have hundreds of millions of people every week that use the, the dictation microphone. Okay. And, um, you know, G board supports over like 140, uh, languages around the world. So I think users of G board have come to expect that it, um, you know, we've been, we introduced glide typing. We have the, the, the mic in so many languages. So I think they've come to expect that we're going to continue to involve, evolve input, um, in a, in a big way. So I, I, I hope this is exciting to them. I hope, I hope people use it and give us a lot of feedback on it. And, um, and we've also made it such that if you don't like it for whatever reason, you can just continue to use the dictation use dictation.

Jason Hiner: You can switch before and get between. So it's just about, Hey, exactly. If you're really good at this, you've already changed your mental model and you're already doing it. You can keep doing it that way. I do think most people will want to use Ramble.

Sameer Samat: I hope so, but we'll see.

Jason Hiner: I will say it's the thing in Android 17 that I'm most looking forward to. It's the thing that comes up in conversations and I help to people who are about using it, uh, are interested in the devices or the, you know, OS it comes up again and again, you know, Samir, when we think forward, sort of these things, the models just keep getting better and better, right? The AI keeps getting better and better. We think about even just in 2026, how much, you know, you mentioned people with boom mics, you know, talking into their computers. Um, I see that a lot too in Silicon Valley. There's whole, you know, floors of people just whispering of programmers whispering into the computers, you know, um, now you think that you think about this idea of us being able to, to talk to our AI agents, to the agents that care, carry out things for us. This has the potential to really change, not just the computing paradigm, but really the work paradigm, you know, in some big ways people so far humans.

Sameer Samat: So here's my question.

Jason Hiner: So, so far human beings, um, have been judged by the work they do, right? The things they carry out. And some of that is like, how good are you? How efficient are you? How, you know, do you put in extra hours? You go the extra mile. If these AI agents can carry out tasks for us and better and faster in many cases, which is sort of what we're seeing, does it really change what is lower philosophical question, what humans are, you know, judge that, what the way we judge workers, if the AI agents carry out things for them, you know, and you think about, you run a team too, in addition to the many other things you do, um, does that mean that we end up like judging or evaluating or, you know, um, humans based on the dots they connect the sort of the goals that they give AI, AI is carrying out more of the work is, are those some of the things that you all think about if the AI continually is going to be better and better at carrying out some of these goals for us?

Sameer Samat: It's a really good question. And I don't think it's an answered question right now. And if I had the software industry or really, frankly, in almost any industry, I think it's still evolving. And, you know, what I probably thought about this six months ago has changed to now probably changed again in C2 in six months, but I definitely think that this, these are new tools. And I definitely think there are areas where they're super useful and there are areas where they're still not as useful as they should be. Um, and I think that when we are looking for team members, we definitely want folks who are, um, just like forever before this, people who are excited about and embracing the latest tools that help them be as productive as possible. Like, yeah, that's just kind of a age old thing, right? You know, and so we look at this as another tool and, and if you come in the door and you are well versed on it and you're ready to go and using these tools for the places where they can be used, um, we think that's really good. And we, we look for that, you know, um, at the same time, we do believe that, you know, at least I've seen in my team, um, that there are places where, uh, the, the tool set is still evolving. And so, um, you know, I think that the, the human oversight, um, the human in the loop component of this is totally vital, you know, and, um, and, and really important, and so it's a mixture right now, but we're definitely trying to give in our team, anyway, we're trying to give people time in their job to learn these tools. So like, if you've been working in one way for a long time, yeah, Google engineers are always curious, you don't have to convince Google engineers to be excited to use new technology, like that's the nice thing about working at, at, um, at Google, but you still do need time because, um, you're learning something new for anyone. It just takes time. So yeah, we've been trying to carve out time for people, um, to, to, to sort of get involved with this and by the way, it's not just software engineering. Like I think product, um, product design, um, product, product management, um, user experience, I think all these disciplines, um, I met with lots of folks across the company who are using these tools in an integrated way with their existing workflows or creating entirely new workflows who are coming to me. And these are my most favorite conversations, you know, where someone says, Hey, look at what I did. Um, here's how I changed how we do this, you know, and no one told them to do that. They figured that out and they're finding it really useful. I think that's super cool. So right now it's another tool. Um, and in the future or six months from now, I bet we'll talk again and who knows.

Jason Hiner: Yeah. What are some of your favorite examples of the way people are working so differently now with some of these tools and what's capable?

Sameer Samat: Well, one thing, I mean, that I think a lot of companies have done is, you know, we have, um, we have a lot of bugs that come in from, um, external and internal sources, and I don't really know of anyone who studied computer science, who was like, when I grow up, what I really want to do is bug tree ops, so she was like, yeah, you gotta, I want to do bug tree guys. I want to figure out if this is P P two P one P zero, you know, we, and then, you know, our, a lot of folks on our team, we have like an on-call rotation, you know, and so people take, take on call rotation and then the bugs come in and the triage and what have you, some of our teams, um, now have built, uh, automatic bug triage agents, you know, and, um, and I think that the interesting thing is the hardest part they've told me for some of that was actually not writing, you know, or not doing the prompting or writing the code around the agent to make all that happen and hooking it up to their existing systems that, that part, I think they, they found was like they could get through that pretty quickly. It was more writing the evals to understand whether or not the bug triage agent they had created was sufficiently good to say, I don't need to have an on-call rotation for this anymore, you know? And, and so that there's interesting that like writing evals and doing measurement of that kind is not necessarily a skill set. It's not necessarily something that everyone has done. Um, and, and so I think that was a new skill for some of those teams to learn. Um, and it led us to also realize that we probably need to package up how to do that properly and make that like something that we push out to folks as like, if you want to help, if you want to build agents to do stuff, then you should learn how to measure whether or not they're good enough, you know, as a skill that people should have. And so I thought that was, I thought that was really cool. Of course, nowadays, when I'm going to a lot of our product review meetings, you know, we always say in the Android team that, um, demos are better than decks. Um, and so it's really cool to see people, um, who are not software engineers come in and present demos, um, prototypes of ideas that they have for what they want to do. And so, you know, UX team, the other day that came in and instead of a user design team that came in the other day, and instead of being like, here are mocks for how we think this should work. They, they have like a functional prototype that you could play with, you know, and that means last something that you would ship, but it was, it allowed you to experience the interactions and it allowed you to experience the, the, the animation, it allowed you to feel what this thing would be like. And we could have a much richer conversation always, right? With a prototype then, but you have with, uh, um, with, with mockups and they did that zero engineering, um, you know, usually sometimes we, we have, um, we have long lag time, you know, because we will want to prototype something, right? So it's shortened that lag time and we can do like five or six more prototypes in that same amount of time. So these are some things that I think a lot of companies are finding these, these kinds of wins super helpful for us.

Jason Hiner: That's amazing. So it's sort of like in, in movies, you could pre-visualize or storyboard before you did it and you wouldn't, those, those things never would make it into the movie. Cause it was like a cartoon drawing of something, but the full vision of what you're talking about was much more accessible to everybody else.

Sameer Samat: No, no.

Jason Hiner: Um, so now you can have product managers or project managers who could vibe code and say, here's what we're talking about. Absolutely.

Sameer Samat: I think, I mean, I, I grew up in Google as a product manager and, um, I was an engineer before that, if I grew up as it was product manager in Google. And I think that right now is a really interesting time to be a product manager, you know, because, um, you're able to express your ability to express where you think the experience should go, I think has been dramatically multiplied by these, these tools that are available. You know, you don't just have to write quote unquote, the PRD first of all, you can have help writing the PRD synthesize tons of information way more quickly, um, and you can prototype and, you know, we, we've even had, you know, we've had a couple of product managers who now are, they're submitting CLs, you know, and still engineers. Uh, but like, that's really cool. You know, that's really cool to have people be able to do that. It's, I think for them for product management, I think it's like an empowering, it's an empowering thing. And, and, um, and it's exciting to see.

Jason Hiner: What's the best thing in AI agent has done for you? Oh goodness.

Sameer Samat: Uh, the best thing. Okay. Well, there's several things. Um, one thing I did with, with Gemini spark recently, um, that was really helpful. And this is a small thing. So I don't know that it's the best thing, but I just want an interesting thing, an interesting thing. Um, you know, I travel a lot for work and, um, I don't, and, and as a result, like sometimes when I book stuff, like I also, I also cancel a lot of flights. And I've got like a ton of emails, you know, airlines send you these emails with your, like your travel credit. And like, I have no idea like Southwest. I have so much travel credit. I don't even know. Yeah. And if sometimes, and I'm booking flights and I'm like, I think I have a United credit, but I don't really remember share. So I asked spark to go through my, all my email and find all the credits and then go to the websites for the airlines and validate that those are actually credits that are still usable at how much is there and when it expires. And it may be a spreadsheet of that. And I was, that's pretty cool to use this. Oh, it's a small thing, you know, but like doing that probably would have taken me, I don't know how long, 20, 30 minutes, you know, only enough that I wasn't going to do it. I haven't done it yet anyway, unempirically. So, um, I think that was, that was super helpful.

Jason Hiner: This is a good, um, this brings up a good point because it'll really do that. It had to have the logins to your, to Southwest and United and all of that.

Sameer Samat: And this, I guess it didn't cause they, they have website, they have a webpage for each of them where you can like go and enter your credit and a pin that it gives you on it. So in this case, it didn't, but I know where you're going.

Jason Hiner: Okay. So the con they thrive, agents thrive when they have more context, they can do things for you when they can do things on your behalf. And so how do you, uh, as Google, how do you convince users to sort of give the layers of context that are going to allow the ages to do so much for you?

Sameer Samat: It's a really good question. Um, I think that, you know, it's, it's, it's really important to start, um, in ways that build trust and have the right transparency and control in place. So we talked about app automation before. Yep. Um, so we designed a few things. So on the Android operating system, now we have, uh, we built something which is effectively a virtual window. So what happens is when you ask your agent, when you ask Gemini, for example, to go and, and do app automation on something, and let me, let's take an example so we can talk through it. Um, I use this example of something I did back around the, the, uh, Memorial day when we were having a barbecue, I made a list of, of, um, people who come to the barbecue, how many people were in their party, whether or not they were vegan, vegetarian, or ate, eat everything. And, um, and then I asked, um, Gemini to take this list and go to, um, uh, uh, uh, Instacart and put together a set of things in my cart that would, that I would need for the barbecue and like, you know, all the burgers and everything else. Right. So it looked at the list, it analyzed it and it starts doing that in its. So let's walk through how that happens. Okay. So there, so, uh, Gemini takes that, understands what you're trying to do. So it understands that you've given it permission to start this automation task in an app called Instacart. So then it is able to now ask Android if it can launch Instacart in this virtual window. So it launches in a virtual window, which means that you can observe what the agent is doing at all times, but you can also minimize that so that it just shows up in your, in your status bar. And, and you see, you don't have to watch it all the time, but at any time you can tap on it and it expands and you can see exactly what the agent is doing. So nothing's hidden from you. Um, and, but if you want to use your phone to do other things, because, you know, the phone is kind of a single user device. It's not really meant for two people to be using at the same time, rather than it can minimize, so you can, you know, browse your email or browse the web or whatever at the same time. And we've also made it so that the agent, um, does not complete transactions for you today. Um, so for example, when it, by the way, if it has questions, it'll ask you along the way, but then when it gets to the transaction screen in the car, it calls you back and it's like, you know, can you look at this and review it? And that's not because we couldn't make it push that button, but we think that's an important safety measure right now to help people build confidence and, and to get, give us the feedback as well. And so I think that, uh, oh, and, and finally the agent is sandbox. So it is not able to, cause you've only said Instacart, it's not able to like go to four other apps either or access the rest of the system. So that, those are all things that we've done to help fight transparency and control around, around this. And that will hopefully build confidence with people over time. And we're getting a lot of good feedback around it. And over time, we, we believe that there needs to be a more robust, even in this permissions model around how you allow ages to do things. Um, and we'll build that out over time with, with developer and user feedback.

Jason Hiner: That's, that's good to hear that last part, because, um, we've already seen, you know, that these models, they are necessarily creative, right? If they can't do something, if they went to Instacart and they're like, Oh, it won't, let me do this one thing. Like an agent will say, well, what else could I do? Right. And it'll sort of, that's where it leads to sort of sometimes jumping the bounds of what it's in. So by having it in the sandbox, by having this vision of we're going to create a permissions framework around agents so that they aren't going to do things you don't want it to do, they're only going to do the things you want it to do.

Sameer Samat: Yep. These things are easier said than done. Um, I think that we have, um, you know, you have it, you always have a trade-off in these, in the creation of these sandboxes between, um, between safety and control and, and in some cases utility.

Jason Hiner: Sure.

Sameer Samat: And you have to find the balance point.

Jason Hiner: We've dialed it.

Sameer Samat: We've dialed it to a point where we want people to feel safe, you know, like, cause all this is very new. Um, and over time, you know, I think people will say, well, why can't I do this? Why can't I do that? And, and I think that's the moment where you start to say, okay, let's introduce additional capabilities, but with the right controls around those and you know, kind of, you kind of, but from there now, of course, there are other instances, there are other, um, and this is like, you have to, when we think about it this way, like you have the phone is there's 3 billion devices around the world, you know, uh, and plus, plus all the other kinds of devices. So, you know, we have to be responsible about this and do it that way. Um, I think there's a lot of experimentation we do internally, um, you know, where we are trying all kinds of stuff and can new form factors and new types of devices and, and, and taking some of the safeties off and all that. So I, I also, I want to balance with like, when you roll things out at large scale, you want to be responsible and you want to do it in the right way. That doesn't mean internally, you're not trying new things and, you know, innovating and I think balancing both of those two things is real.

Jason Hiner: And there's some user education that goes along with that. If, if, if an agent can't do something and then you prompt user, would you like it to be able to do this, then you sort of learn our learning. Oh, there are these sort of controls in place. Am I comfortable removing or reducing sort of the level of those in order for it to do more things. So that's part of the journey that we sort of have to go on in the time ahead.

Sameer Samat: I think that's right.

Jason Hiner: Okay.

Sameer Samat: Okay.

Jason Hiner: How about you all have other devices as well. You've re-entered the smart glasses case, so to speak, there was Google glass, you know, a decade ago, which, which was, it's really interesting to feel how far ahead of its time it was, you know, at this point, really interesting. What does that look like? Cause you start to think about, uh, devices and the devices going beyond the phone, the watch can do some things. Um, the glasses now can do some things. What is Gemini intelligence, um, look like, what is sort of the intelligence system kind of look like as it spreads beyond the phone, unless people do things that not only, you know, they are maybe better or different, but in some cases really, they have never experienced yet. Yeah.

Sameer Samat: I mean, I think that the, some of these form factors lend themselves to this kind of different human came computer interaction, even more than the phone. I think the phone is become this, um, highly, we're all very experienced and we're highly tuned to like, use our phones to do stuff pretty quickly. And so yes, there are like more multi-step tasks and what have you that you might want to do, but like, call me a car. I can do that pretty quickly in my, my, we're a list app, you know? And so the bar is really high for these things to be super useful there. Um, now if you take, if you take a watch where you're on a run or you take a pair of glasses, where you don't have your phone out and you don't, maybe you don't have a screen, um, or you're in a car where you really shouldn't be using your hands, putting any except drive, you know, I think there, um, the, this new voice understanding mode where you can get things done, um, in an automated way, I think is going to be far more appreciated quickly. Um, you know, if you're on your glasses and you can, you walking down the street and you can say, Hey, you know, order my regular from whatever cafe, you know, is a mile away, um, and it can do that for you.

Jason Hiner: It's pretty useful. Yeah.

Sameer Samat: You don't have to take out your phone from your, your pocket, your purse, and you just got to continue on with your day. Um, same thing in the car. You know, if you're, if you're like, Hey, you know, um, could you put together a quick like audio overview of what I need to know for this meeting. I'm about to have when I reach work in 15 minutes and it does that for you. And like, you know, I mean, you could probably figure out how to do that. I pull up notebook, LM fumble around, you know, do things where, um, on your phone, but like you shouldn't be doing that while you're driving.

Jason Hiner: Yeah.

Sameer Samat: And so I think this is Gemini intelligences is of course about the phone, but it's, it's about so many more of these devices and how they all work better together. Um, and I think that really helps differentiate the Android experience. Um, you know, in the future, I think that's what we want it to be about. And one of the things I did want to mention around this is how we talk about it with consumers. I am maybe that's just like a briefly interesting thing. Um, I think the, the, the word AI has created a lot of feelings. They say, yeah, he's right. And, and if you, if you dissect that, there's so much going on there. Which we don't have time to talk about, but I think most importantly is, you know, for, for folks who are in technology and, and, you know, I, I, I studied computer science and so I'm, I'm, I'm really excited about what's going on with the technology and can talk about that, you know, kind of endlessly. Um, I think for folks who aren't really as interested in that, maybe they have other passions in their life besides that, I think the word actually is a little scary, more than a little scary. I think it kind of turns some folks off. So one of the things that we're really trying to do is just talk about the benefits of the features, you know, what, what does this do for you in your life that actually will help you? So it's even if you never mentioned the word AI, actually I try not to, you know, because I find that when I'm talking to, to folks who are not in the technology space, it helps them really understand why I'm excited more than if I start using the word AI because here we go with the every go with the AI, you know, and so, so when I talk about, you know, things like circled search on Android, or, you know, I talk about what Gemini can do with automation, whatever I talk about how like, you know, you saw that necktie on that person or that, or that purse that you wanted, you can just circle it. You could never describe that thing to Google. Cause it's like, how do I describe this? Yeah. This, this, this thing, you know, but now it's like, it can show you it, show you how to buy it and people are like, well, that's, that's really cool. Tons of AI behind the scenes.

Jason Hiner: Share, right?

Sameer Samat: That's our scene vision and everything. Um, and same thing with automation. Um, you know, when, when people are here, some of the examples and, and, and how it can save them time, they're like, Oh, okay, that's cool. You know, now you're, now you're talking. Um, so I just wanted to share that as well. That's kind of like a, a bit of all of this. You mentioned Google glass, right? A lot of all this is like, we are human, right? So like we want it to be socially awesome and also useful, right? It can't just be one or the other, right? Um, to, to truly be a, to, to be a hit technology product.

Jason Hiner: So more human centric than technology extension. I think so.

Sameer Samat: I mean, with glasses, for example, one of the key learnings, obvious in retrospect is like, it has to be something that you really, really want to wear. Yeah. You know, um, you know, it has to be something that like a good test is like, if the battery, if you forgot to charge it and the battery was dead, would it still be something you would want to wear? And if you're like, yeah, I like it.

Jason Hiner: It's nice.

Sameer Samat: It looks, people compliment me when I have it on and you're like, it's a good quite, okay, now, now we're getting somewhere, you know, now make it useful. Also, not just, uh, you know, a, a fashion, uh, accessory. So I think that's, that's a lot of what I think we have to get.

Jason Hiner: Right. Google does so many things. Like we've talked about a bunch of these things. That is this great benefit because what we talked to earlier that, you know, you don't have a one experience on your computer or different experience on your car, different experience on your watch. You know, if you were in the AI world, you know, it gets super frustrating. If you're like, you talked about something with your AI earlier or set a task, and then now you sort of want it to, um, adjust that, or you want to do the next step and oh, it doesn't have context for that, or it doesn't know that the, the flip side of the coin is because you're Google and you do so many things, it can get pretty confusing to, to users, like the branding of different things, the names of different, um, things. How do you navigate that? I'm sure it's one that you're thinking about because you have a whole portfolio you look at, you, you and your team are trying to connect the dots across these experiences. Um, how do you think about that in terms of the brand and how the user interfaces with these different pieces? So the user in some cases doesn't have to think about it, but then also doesn't get confused about it of like, Oh, there's Rambler here, but it's dictation here. Is that the same thing? Does it, does it know me? You know, that I assume is something that, that you have to think about a lot. Yeah.

Sameer Samat: Uh, definitely. And, and I think that, you know, we want to try to make this as, uh, as simple as possible, of course we are coming up with new things all the time. And so, you know, we do have a user base that wants to try new things and, um, they want to refer to it as a thing. And so, uh, but as you scale these things, you need to keep it simple. Um, and so we try to manage both of those two things. Um, we, we often, I think look at it from the standpoint of, um, if we are going to explain this in 30 seconds to, to, to somebody in an elevator, you know, what, how would you, how would you do that? And there are definitely instances where, where internally before we are going to launch something, well, much before we're going to launch something, we, we experienced it like, okay, actually that was a little too complicated. And we, we try to simplify it and we don't always get it right, but you know, we take a lot of feedback and we, we iterate on that and try to make it easier. I think one of the nice thing, one of the things that I think is a positive about Android is it is an open system. And there are a lot of OEMs and software developers and what have you that are building like really cool stuff. And so our users do have a little bit of an expectation that, you know, on Android, you always see the future first. And, and so they do come there looking for that, for that future. So we do, we do talk about some of those things and point people to some of those things, but I think when we do that, we're also talking hopefully to a particular audience that really appreciates that. Um, and then when we look at like the broader audience, if you will, you know, I think it's really important for things to be simple, straight forward. Um, and, uh, and very, very, very easy to name and understand.

Jason Hiner: If the, with the pixel 11, if people start calling it the first agenda phone, are you happy with that?

Sameer Samat: Oh boy. Um, I, I mean, I think if they're talking about it that way, that means we're doing probably something very interesting. So that's good. Um, there's so much in these words now. I don't, I don't even know. Um, but what I would say about the pixel 11 is that it really embodies a lot of, it embodies what was the tip of the spear for where we think, you know, the Android experience is going. And, and if people are excited about it, um, and they think that we're doing interesting things there, um, then, uh, you know, I'm, I'm really happy with, uh, for what that means for Androids.

Jason Hiner: What's the thing that, uh, you can't wait for it to do, you know, that we talk about, uh, agents a lot and the genic things, and there are a lot of things that can do that it couldn't do 12 months ago, maybe six months ago. Um, what's the thing that you can't wait for, you know, the phone.

Sameer Samat: There's so many things we're working on. Um, I mean, one that I can share is I do think that, um, uh, um, again, I'm going to make it very practical.

Jason Hiner: Okay.

Sameer Samat: I'm not gonna, this is not going to be something that, you know, changes the world, but I'm just going to make it practical because I think it's, it's not going to be far off, you know, this is like next few months kind of thing where we're going to start to do this. Um, I hate filling out forms. I'm tired of them. I want them to go away, you know, and, um, you know, today, when, you know, that experience, when you hit a payment flow in an e-commerce website and it just fills stuff out for you, you're like, thank you. And we just all are used to that now. Right. But, you know, if I have to remember my kid's passport number, like one more time, I'm going to lose my mind, you know, and, um, and so I think it's 2026. We have all this technology, like why is it that I'm still running around, you know, with, with, you know, keep notes of things that I need to cut and paste and put places or it's in my email somewhere or what have you. So we're working on a, an expanded autofill system that with your permission will be able to, to do things like, you know, what's your TSA number and what's your, what's your passport and, you know, these kinds of things that you just, we all look up, you know, I don't want to do that anymore. Um, and I think that's a very practical thing that I, um, I'm like super excited

Jason Hiner: about. That's a great one.

Sameer Samat: I love it.

Jason Hiner: All right. I always ask the same two questions at the end of the podcast. The, the first one is what's the tool that you're using right now? Um, it's an AI podcast. So it was AI tool that you're using right now that, uh, that you think is maybe overlooked or underrated that you would love for people to, um, to know about.

Sameer Samat: Yeah, easy Gemini notebook. Okay. And I still find that, um, folks know Gemini notebook best for creating these audio overviews of stuff.

Jason Hiner: Formerly notebook alarm.

Sameer Samat: Formerly notebook alarm. Um, and I think it's great at audio overviews use that all day long. Uh, but, um, an example, my, uh, now sophomore was freshman last year was doing finals and he's like, dad, can you help me with, with the science final study the stuff I totally know it. I'm sure. And, um, and so it came, I was on my way home that, that day. And I got his, I got a syllabus. So I have any of this stuff.

Jason Hiner: It's all Greek and they're like, Oh my God, like I'm going to need to like, I

Sameer Samat: mean time, but like, I gotta get home. He's going to be like, we're studying study hour. Now he's got the test. And if he did, so I put all of that into Gemini notebook and I told it, and I asked it to make a video overview. Okay. And so this is like, it makes a full on and animated presentation of the material. It took the chapters of the syllabus and it literally walked through each one and was like, okay, we're going to talk about tectonic plates right now. And then it like illustrated them and the different things that happen and all this kind of stuff. And he and I just watched it together. It was like 30 minutes. And like, we can, I watched it together and, and he's a visual learner as well. And I know a lot of people that are. And, um, and so for him and for me, this was a lifesaver on, um, and it did that in about 20 minutes on my ride home. I put it in before I got in the car and then afterwards. So I think that tool for learning, um, and for, um, for, for, for getting up to speak quickly on a subject is invaluable. And if you're not using, you're not using Gemini notebook, like check it out. It's like really, really cool. That's my favorite one.

Jason Hiner: It's very good. It's also, um, I find it is like a, uh, kind of a hallucination proof, almost, you know, solution in that you can essentially say, I have this topic. I'm trying to learn about, I'm going to give you all the sources.

Sameer Samat: So it's grounded on all that.

Jason Hiner: Yeah, to, to ground it in. And then I'm going to ask you questions about it. I'm going to have you organize information. Yeah, it's very powerful.

Sameer Samat: And by the way, it also has a mode where like this is built in. You don't even have to prompt it. There's a button you can press for this video thing that I talked about, but there's another button you can press for build me a pop quiz.

Jason Hiner: And so like we, so we did that afterwards too.

Sameer Samat: And I was like, okay, make me a quiz. And then it like gave him like a quiz, like a practice test, basically. And like, and it has like explanations for, well, you can ask it like, why is this not right? You know, and it'll like explain it to you. So I found that to be super useful. Very cool.

Jason Hiner: That's a great one. All right. Last one is, you know, the promise of the AI has always been like, Oh, eventually it's going to do these things. Like you just grabbed with forms. It's going to sort of automate away some of this, this stuff we don't want to do. Right. And it will give us more time. What I find in 2026 is most, you know, leaders, it's not that they have more time. In fact, they're like stretched even more for, for time. And so they, they asked this question of like, right now, how do I get maximum leverage for my time? And so as you think about it, so this is bigger than an AI question. It's more like in the age of AI that we're in where we have all these, we have more capabilities at our fingertips. I think that's why people are working more because they sort of see all the possibilities. Um, what's your best tip for leaders in maximizing their time and getting maximum leverage for their time?

Sameer Samat: I mean, I w there's two.

Jason Hiner: Okay.

Sameer Samat: Um, so first is you've really got to look at, um, this for me, I, I really look at what, um, where am I spending, uh, the empty calories? Okay. So, um, as an example, uh, on Friday, for whatever reason, everyone wants to send me their, their like summary of the things for the week. Right. And, um, and it's always like 5 0 1 PM. This like, I one time printed out all this stuff, uh, for Stephanie ahead. I said, y'all sent me about a hundred pages of stuff. Like, and like, I don't need you to send it to me, but they want

Jason Hiner: to let you know what's going on.

Sameer Samat: So, uh, reading all this is, is great. And I appreciate it, but it also is a lot of time. And so I call this like, you know, there's, there's, there's chunks of time there where you just have to identify where you can use, um, this technology to help automate, synthesize, et cetera. Now that you do that. So now you've done that now you using all the tools to your best ability. Right. Just as you're all trying to do right now. So I'm here, here's my other tip that is completely independent of the AI era. Okay. Um, and I call it a calendar cleanse and, um, anyone who's ever tried, um, to make time on their calendar for their most important things. So first you make a list of your most important things. Yeah. These are subtly not the things that are only the most important, but in an organization, they're the things that are important that uniquely require you.

Jason Hiner: Okay.

Sameer Samat: So they may not be your biggest goals. They may be the things that if you don't help with things won't happen. Maybe it involves two different teams, or maybe it's super complex for this reason or whatever. Um, so you make that list on everyone. Let's say it has that list and it's you make it as short as you can, but it's always a little longer than you want. Now, anyone who's ever experienced trying to go to their calendar and like, okay, let me see what I can delete to make time for this. That I find to be mostly an exercise and like, as a folly, right? Because you go through it and you delete like two meetings and you, you, you add a bunch more and then you're like, I, so we do a calendar cleanse every quarter. We make the list of the most important things. And then my, um, I have an admin. So my admin and I, we sit, we sit down and we, and she hates this process by the way, we, we delete the entire calendar, just select all delete. And, um, and then we rebuild the calendar with as much time as is needed for each of the things on the list. So we go to the first one. We're like, what are the things we would need to do here? That would make this successful. Well, okay. I would probably need a weekly on this. I probably need to meet with this person regularly. And we, we put that all on the calendar and the times that they should be that are optimal for getting this stuff done. And when we're done with that, we have whatever time is left on the calendar and we go back to everything else we deleted. And we try to put what, what the most important stuff is in there. And I do that every quarter. Um, and that's the best way to get the end of the quarter. You're like, okay, I, I, I actually spent time on the most important things.

Jason Hiner: That means that's the most useful and specific answer ever for that question. I love it.

Sameer Samat: It's, it's painful. Uh, like when you do that, when you do the select all delete, you don't look. Just do it. Because if you start looking at the things, you're like "But I can't. But what about this one thing."

Jason Hiner: Very good. Samir, thank you for the time. This is a lot of fun.

Sameer Samat: Appreciate it.