How to Build a Beloved AI Product - Granola CEO Chris Pedregal

AI is going to let humans work differently, think differently. There needs to be a tool that supports that and that's what we want to build. So this idea of a contextually aware workspace, like AI powered workspace, like that's what we wanted to build with Granola and we said, okay, great, perfect, we got the vision, that's what we want to build. Where the heck do we start?

Hi, I'm Matt Turk from firstmark. Welcome to the Matt Podcast. My guest today is Chris Pedregal, the CEO of Granola. In just over a year since it was launched, Granola has emerged from the crowded category of AI note takers as a bit darling in Silicon Valley. Not just as a hot AI startup, but as an AI product that many in tech circles use rapidly every day and often describe as life changing. This episode is a masterclass in how to build a beloved product in the age of AI. Full of practical tips and lessons that Chris has learned along the way, including how to achieve simplicity in product design.

We looked at it all and we cut out 50% of it. We basically redesigned and cut out 50%.

Knowing when to exit stealth and it's.

Really, really busy market. Launching something more polished so that when people use it, they're wowed by it is a way to stand out and.

What it's like being a lead entrance to a category and then having to compete with a bunch of big companies, including potentially the OpenAI's of the world.

OpenAI is going to try to do everything to everyone and I think the question is, can we do something way better for a specific use case and specific type of user?

There's a lot to learn for AI builders in this one. Please enjoy this great conversation with Chris. Hey Chris, welcome.

Thanks, Matt.

All right, so not to fanboy you from the very beginning of this conversation, but I have to say I'm a very rabid user of Granola and actually our entire firm at firstmark is. And you know, when I started using Granola a few months ago, I thought I was pretty cool, a pretty early adopter kind of situation. And then there was this article in the Information a couple of weeks ago that basically said, well, everybody in Silicon Valley uses the product all the time, so maybe not so much of an early adopter from that perspective after all. Just curious about how that feels as a founder to have a product that's just widely embraced by our entire, at least little tech ecosystem.

It is, it feels both amazing and daunting, is the honest response. I we did this. This was last maybe November So we're based in London, right. And we went to SF for a board meeting and someone on the team said, hey, should we rent out a bar and just email users and say if you'd want to come and we'd like, sure, you know, and we thought, we thought like five people would show up and like this two floor bar was just full of, full of people and, and then they were just the level of detail with which they were talking about the product or things we should change or things that they had noticed. It really hit me because when you build a product for people, but really being in a room with that community all at once, it made me realize, oh, there's something special that's happening here that we didn't necessarily design for. It's kind of organically happening and now it's kind of our job to follow that.

Yeah. And one amazing. Which I'm sure you've heard tons, but it's just my personal experience and looking online and talking to people. A description of the Groanla experience that keeps coming back is life changing, which is insane. But that's again, true in my experience. I tweeted that at some point I was, you know, all my life a rabid note taker. That's how my brain works and it helps me think through the meeting or whatever I'm listening to. And pretty much overnight that lifelong habit just disappeared. Once I tried Grinois a couple of times and trusted it, which again, not to fine bore you, but it's been an incredible experience.

I appreciate that. I think that speaks to the moment in history that we're living in where AI now has all these capabilities that are going to transform the way we work and the way we think. And hopefully that experience you just described with granola will keep happening in many more aspects of your work life as your. You're basically able to outsource lower level tasks and allow you to think higher level, which is why I think it's such an exciting time to be building and to be living, quite frankly.

I think when you released your team product a few months ago, you used the expression second brain and it's basically what it feels like. And in some ways a life changing part does have a little bit of a daunting aspect as a user because it sort of feels like you're out outsourcing your memory to technology. And memory is such a part of who we are as humans and that's how humans survived for centuries. Whoever was able to remember facts was able to just function well in society. So it does feel like a wonderful tech journey, but something a little more than that, actually. Possibly.

I completely agree. And I think there's going to be. As AI gets smarter and we build more and more tools and workflows on top of it, we're constantly going to be letting go of things that we used to do and letting machines do that for us. And there are times when I think that's net. Incredibly beneficial, and I think there are probably times where that's harmful, right? Like negative beneficial. The example I always use is Google Maps right on the phone. So it's very clear there were cities I lived in before Google Maps came out, and, and I can go back to them. I can still navigate them without a map. And they're the cities I've lived in since Google Maps have come out. And I have, like, there's a very small area of that city that I can navigate without, without a map. So you could say, oh, my, my navigation skills have atrophied 100%. The number of hours I've walked around lost in a city trying to find a place has also gone down dramatically. Right. So it's like I'll, I'll take, in that instance, I'll definitely take that trade off. But I think you have to be thoughtful, case by case, about what tools you use and what you decide to outsource and do.

We think term that what happens, like, we become. We have more time to actually think and reason, but it sort of feels like models are doing that for us as well.

I mean, I, I think this is the, this is why I'm so excited to be building Gridola right now, to be working in the space, because I, I think that future is kind of up to us, right? In terms like there's this great quote which is, we shape our tools, and thereafter our tools shape us. And when you think about AI and the future world and how it fits into our society, I think there's this big question of, like, what, you know, what do we outsource to AI? Where does AI replace humans and where does it augment humans? And we're like, me personally, Sam Granola, we're really a big fan of the augmentation idea. This goes back to Douglas Engelbart in the 50s, right? Augmenting human intelligence and his view, honestly, his stuff, I feel like people don't talk about him enough. It's just so inspiring. He's known for being the inventor of the mouse, and I think the mouse is like the least important thing he's come up, up with. And basically this was when computers barely existed. Like the computers that existed were in the military and the Navy and they, they like filled up whole, whole floors, right? And there are a bunch of folks actually who are true visionaries at that time who imagined this world where computers would be accessible to people, they'd be common and they would be tools for work and tools for thought. And the way Engelbart talked about it, he was like, we are becoming, you know, the world's becoming more globalized, the world's becoming more complex and we need better tools to help us collectively solve more complex problems. And that's, I mean, talk about an amazing narrative, right? I feel like we don't get enough of that today. You know, I wish, I wish whatever. I, I have a lot of love for, for the tech world, but you know, there's something to be said about some, you know, counterculture builders in the 50s, 60s, 70s, even like Steve Jobs when he started Apple, like these folks had like, you know, a counter revolutionary world view of the world and like how personal computing and tooling could be used for that. So in terms of the atrophy, the future, like, I guess I was trying to think about examples of this is, have you seen Wall E, the Pixar movie?

Yeah, yeah.

You know, the humans, like, the humans.

Are fat, can't walk, people floating in space.

Exactly. So I feel like that's one extreme future for humanity post AI, right? And then there's another one which is I think maybe like Jarvis from Ironman, right. Which is kind of like, okay, now I can fly, I can solve things I can never do before, or am I like this fat blob floating through space? And I think it's a little bit what tools we build, what bets we make as a society, what rules we make. I think that's going to be decided.

Over the next 10 years from an entrepreneurial journey perspective. One of the parts of the granola story I find fascinating is that you guys were kind of late to market. In many ways, the idea of an AI notepad is not new. There were several companies doing that. There were large companies like the Zooms of the world doing that. So for the builders listening to this, who may look at a category and see a few companies and try to decide whether they should build in this category or ignore it and find a category which is less crowded, you guys think about, oh, we can come up with something that's going to be better than all of them.

It's a great question. So I think the answer really Just comes down to like what were we trying to build? When we set off to start Granola, there have been like meeting transcription or recording products like Otter and Fireflies I think are like nine years old, right? They've been around for a long time. Like that's not a new idea. There are all these tools that are like, okay, we can now record meetings, right? And try to make something useful there. That's not at all where we started with Granola. We don't think like a meeting recorder. That's not what we're building. We want to build a tool for thought. Like the genesis of Granola was I quit Google because Google bought my last startup. So I quit Google knowing I wanted to do a new startup. And I came across LMS for the first time and they blew my mind. And I was like, this is going to change. Everything is absolutely going to change. The tools we use for work or productivity tooling. And I met my co founder who had come from a Tools for Thought knowledge management space. And we basically said AI is going to let humans work differently, think differently. There needs to be a tool that supports that and that's what we want to build. So this idea of contextually awareness workspace, like AI powered workspace, like that's what we wanted to build with Granola and when. And we said, okay, great, perfect, we got the vision, that's what we want to build. Where the heck do we start, right? And that was like, oof, okay, well kind of, kind of imagine you can imagine an assistant that like knows everything about you, is there where you're working, gives you suggestions, learns from you. You know, you can kind of imagine that. But where do you start as two people you start know building in 2023 and we, we realize that AI is only as helpful as the context it has about you. So this is something I think we don't talk about enough Even today in 2025, like context is so important. Context, design, curation, like that's a whole topic. Maybe we can talk about Matt. But so it's like, okay, to be helpful to a user, we need to have their context. And as a tiny startup with, you know, like we need an entry point, it kind of came down to email or meetings. Those are the two places where there was a lot of useful context that we could access. And then you put the product building hat on and you say getting someone to change their email client is hard, right? That's like a very, very tall ask. Whereas taking notes in meetings, honestly, the biggest competitor that Granola had From day one and even today is Apple Note. It's this idea of, you know, it's like I'm in a meeting, I'm five minutes in, you say something smart or that I need to remember that I'm like looking for a pad or paper or something to write it down. And Apple Notes is the virtual version of that. That's how we started with meetings. We kind of begrudgingly entered this super saturated space, but we really did, we really did think about it as a very differently, I think, from the companies that were out there. Like we were thinking about granola as a personal tool for you to help you do your work better. And while there have been tons of meeting recorders out there, you know, I'd posit that none of them feel like that. Like when you log into these, they feel like a meeting repository, like a here are recordings of meetings or just the fact that a meeting ends and it emails generic notes to everybody that it was in that meeting. It's a completely different feeling than like, here's this, here's this tool for me that is optimized for me. Like I was. To be perfectly honest, I was surprised that we were able to break out in such a crowded space. Just like there's so much noise, there's so much happening, there's so many people doing things. And granola is by design very quiet. It doesn't. There's no like growth hacks in there. So that was, that was a really pleasant surprise.

Amazing. And you mentioned your prior startup and your co founder. Another very interesting part I find is that both of you guys are product people, right? I mean, I think you have a computer science educational background, but like you, is that fair to say? Like neither.

That's accurate. Yeah. My co founder is a designer. I'm a product person. We can both code. He can code much better than me, but we're product and design. Yeah.

And where I'm going with this is I'm curious what that means in terms of again for people builders listening to this, what that means in terms of what kind of team one needs to build an applied AI company these days, a company running on top of an LLM. So what was your level of sort of technical comfort working with LLMs? And at what point did you feel the need to start bringing people to do more technical stuff?

I think the main thing that's changed here is that it used to be that you would need really strong technical chops just to build an MVP to understand if this is something people wanted or not. And I think the reality now is that that isn't the case. You can usually figure out MVP or like is there a there, there may be even early product market fit potentially or signs there without a whole bunch of technical acumen. As long as you're building on top of the models like it's a completely different story obviously if you're building at the, at the model layer, but if you're building a, a rapper, a rapper company like, like we are, then it's like you can, you can learn a lot. And in those early phases, like when I was looking for a co founder, I met Sam, but I was also, I met all the, all the LLM experts from Imperial and Oxford and Cambridge because I thought that was DNA we would need on day one. And as Sam and I started prototyping, we realized actually there wouldn't be much for that person to do until we figure out product market fit, until we maxed out on what like the base model, like the off the shelf models could do. And then we would need that expertise and then we stopped looking for that person.

And to the ICL and generally London discussion that you mentioned, it's also interesting and a little bit of a narrative violation, if you will, that you guys are building the company out of London in a world where like the default sort of zeitgeist thing that people repeat to one another is that you can only build great AI companies in Silicon Valley or San Francisco. What has that experience been for you? I guess first of all, why are you doing it? I think I read somewhere it was for personal reasons.

Personal reasons?

Yeah. And then more importantly, what has it been like?

It's funny, I, we were here, we're in London for personal reasons. My wife's, My wife's English. We moved here. I knew I wanted to do a startup. We chose London because there's an amazing engineering talent here. There's enough of an ecosystem here to really have a go at as a startup. And then when I decided I wanted to build an AI startup, I said, oh my God, yes, like DeepMind's here. Like a lot of like modern AI was like invented here. Some of the best programs, like I said ucl, Cambridge, Oxford, Imperial, they have amazing AI programs. And then, and then I had the realization that actually, you know, product and design and just general product taste and building is super important. So we didn't need to hire those folks early on. I think there are trade offs, right? I think there are very real trade offs. There's like a center of gravity of talent In Silicon Valley, I think we're in a very lucky position to be, I'd say, one of the most visible and desirable consumer facing AI startups in London. So for the continent of people over here, they find us, which is incredible. And I think in an era of where taste matters and product sensibility matters, there's just amazing talent here for that as well as amazing engineering talent from all the big tech companies. And there's a huge influx of Russian tech talent that's coming to London. So we're definitely not in the eye of the storm, so to speak. The which. Which I think is, to be honest, mostly negative. I think the, the upside is probably that it's a little quieter over here. There's so much noise, there's so much change, there's so much thrash in AI. Like whenever I talk to AI founders, especially second time founders, they're like, it's never been like this before. Like this is like founding is hard. Founding an AI right now is like emotionally draining, energy draining. It's like draining in every aspect because it's so fast and so, and everything can change, can pivot on a, on a dime. And I think being in London insulates that, insulates us from that a little bit.

What's particularly interesting is that you, you're in London but you're a Silicon Valley darling product, right? Typically the, the trade off is like, yes, you can build great companies outside of Silicon Valley, but typically Silicon Valley ignores you. And I think you probably maybe was lovable like the, or Synthesia, you know, one of the rare companies that has sort of broken through the consciousness.

So I guess that was extremely intentional. Right? So we're, we are, the way I talk about internally, we are an American company that happens to be in London, right? And we built for Silicon Valley, we built for the American market explicitly. If I ever see any copy that has English spelling instead of American spelling that goes out, I throw a hissy fit because I want everyone to think we're an American company. And, and that's what happened. I also have to say like, it really helps. That helps that I built my previous company in the U.S. all our investors, or like our main investors are based in America, had that network I had. Like basically it's like that DNA we've transplanted to London. So it is Granola is a Silicon Valley DNA company that happens to be building in London and leveraging that as much as we can tell us about.

The beginning of the company. So the product itself launched in May of 2024. I believe, which is not that long ago at all, given again the level of heat and love for the product. But I read somewhere that before that you were in stealth or in building mode for about a year. So again, for builders out there, how did you think about when to launch, when not to launch? You know, there's this constant tension between, like building public, you should be embarrassed by your first version, otherwise it means you launch too late. But on the other hand, you only get one chance to make a first impression. How do you think about this?

Two thoughts about this. The first is a simple way to answer. This is, what is the fastest way for me to learn? Presumably, you start building something, you have some prototype, you have some early version of the product, and you say, will I learn faster if I launch publicly or will I learn faster if I don't launch publicly? The answer for us for about a year was we'd learn faster if we didn't launch publicly, because we were onboarding users every day onto granola and it was painfully obvious what was broken about it. So launching publicly and getting 10,000 people telling us the exact same thing was actually going to slow us down, rather than just fixing it based on what users were telling us. There are many costs that come from launching publicly where it's like, now you have users, you can't ship things with bugs. If you pivot, it comes at a cost. We basically spent a year onboarding people, learning what was wrong about it, making fixes to that, onboarding a new set of people fixing it and iterating. And then basically the moment when we said, ah, like we now, we now have something that works and we're going to learn a lot more by having lots of people use it and realize who, who does it take off with, right? It's like maybe real estate people will love it in a way or use it in a way that we didn't expect. Like that was the moment we decided to launch publicly on the general wisdom on this, like mvp, non mvp. I think today there are so many products and companies coming out and vying for your attention that launching something more polished so that when people use it, they're wowed by it, is a way to stand out. So I do think it's tension. You shouldn't be tinkering in your closet for two years and the world's moving very quickly. But there's a lot of, kind of a lot of MVPs floating out there, right? So if you want people, if you want to stand out in this really, really busy market, I Think you need to have something a little bit more polished again when you try to draw.

Attention to it while you were in that building mode, semi stealth. How did you find those first users? So you mentioned you're. You're very deliberately an American company based in London. Were you also very deliberately a company targeting, I hate the term, but like, for lack of a better term, tech elite, quote, end of quote, of a bunch of top founders and VCs. Was that intentional or that that sort of happened?

Yeah, well, it's two stages. One, at first we built. We were building for us, right? And then we were. Then the first users were kind of friends and family and extended network who were knowledge workers, used computers, did a lot of zoom calls, and that got us pretty far. And then there's this moment where users started telling us different things. They're like, oh, this is what's important, this is what's important. And we said, okay, we know granola will be a general product, a horizontal product. Lots of different types of people are going to use granola, but we should just choose a user type on day one to make it really good for and then expand out there. We kind of looked around and we said, okay, we need a user type that has a lot of meetings, relatively formulate, you know, like lots of a similar type of meeting with relatively formulaic note style that they need VCs that we have easy access to. Yeah, exactly. VCs. Right. Yeah. So we said, okay, let's build for VCs. Yeah. And then. And then as soon as we launched, we said, okay, great, now we're done with vc. No, they're not going to be. We're not going to focus on VCs. We're going to focus on a different user type. And we chose founders just because we thought they'd be the hardest. Like if you founders might have a sales call and then a user feedback call and then an interview, and basically thought if we could build a good product for founders, then a great product for founders would be by default a decent product for folks in these other roles. And then we could make it better over time.

All right, so getting into the product itself and the general philosophy of how you designed the product. So the, the key first thing, which to me feels like the killer feature, or at least a clear differentiator, is that decision that Granula should be hidden or at least not apparent to other participants in the meeting. In stark contrast with, as you mentioned earlier, bot first kind of note takers, where you're on a zoom call and Then there's somebody else's insert company name, you know, note taking bot. And I mean, clearly there's a little bit of sort of sensitivity here around confidentiality, privacy. And to me it feels like in retrospect, like a little bit of a certainly opinionated, perhaps gutsy kind of product design decision. Curious about the genesis, how you thought about it from a product standpoint, but also from almost societal standpoint.

We always started from the, we always started from the perspective of this is a tool for you, what will make for a great tool. And there are a few characteristics that are really important. So a tool needs to be consistent and reliable. Like if you pick up a pen and it only works half the time, that's a terrible pen, you're not going to use it. Right. And in the case of meetings, there's this very real thing where some of your meetings or conversations might be on Zoom or Meet or huddles or WhatsApp or, or maybe not even on, on, on a VC. It might just be in person. So we started off from this like tool building perspective of like granola needs to be consistent and it needs to work across everything because we have this like 500 millisecond window when someone is in a meeting and decides they need to take a note, like, what tool do they open? And again, we're competing with Apple Notes and Apple Notes always works. It doesn't care where you are, what you're doing, it always works. So that's where we started. And then from the adding a bot to the meeting perspective, well, like technically, just because we wanted to work everywhere, that wasn't a good option. But if you analyze that a little bit as a tool, bots make you feel kind of weird, right? Just like a big black box on the screen. It's like not a person. Sometimes they show up before you, you join the meeting. It's like a kind of this awkward thing, beautiful thing from a growth distribution standpoint, right? Like you get a user, now they're exposing everybody they're meeting with to your product. So everyone thought we were kind of crazy not, not to do that. And then, and then the way I think about information capture and usefulness is basically, I'm sure that two years from now, three years from now, everyone's going to be using something like Granola. Like I'm hoping it's granola, but if it's not granola, something like granola just because it is so useful and will get so much more useful over time. And I think as a society we need to figure out, like, what are. What are the right norms there? And I think what you basically want is you want something that is the least invasive for the maximum useful, and that's like the. The right trade off here. And when we designed Granola, we basically said, okay, because all. All the other tools out there, they record audio, they record video, they save that stuff. At least when we started off, that's how they. The tools worked. And we said, that's that again, that doesn't feel right. Like. Like I don't. I don't want to be recording. I don't have video recordings of all my meetings. That feels very invasive. Like, what do I actually need? Like, I actually need good notes, right? That most of the time, I actually just need good notes. And so we made another decision early on, which was, even though we could store the audio and that would be useful, we do not store the audio. So we don't record the audio, which completely changes the way Granola feels. I think Granola feels more like a really smart enhanced notepad than a meeting recorder.

You store the transcript, though, and people can review the transcript and query the transcript. Very important.

Exactly. We stored the transcript. We actually were hoping to not even do that, or at least not make the transcript visible. And I guess one of the. One of the things we figured out and has become a design principle for us is that in the world of AI, where AI makes mistakes, transcription makes mistakes, it's really important that I don't have to trust. I don't have to trust the LLM output. I can kind of go back to the source and. And of course, transcripts get stuff wrong all the time. But it's like, if I can see, if I read the transcript, I'm like, oh, that looks fishy. That's important as part of the experience.

Yeah. The user becomes the human in the loop, effectively. Talk about simplicity. My personal experience with Granola as a user is that it's incredibly simple. It's incredibly frictionless. But as we all know, simplicity from a product design perspective is very hard to do. So I'm curious about how you think about it and perhaps what you decide to deliberately not include that would have ruined that simplicity feel.

We had an event the other night, and we were reminiscing over beers about the versions of Granola we built before we launched publicly. And basically, what happens. We were in stealth for a year, and as I said, we were onboarding people every day, learning about what was wrong. And we kept adding things and adding features and adding views by the End there, there is this version of granola where you could, you could kind of swipe and there are all these panels and it's like, here's your transcript, here's your super long, like, here's your blow by blow of exactly that happened in the meeting. Here are your notes, here are your private notes, here's your, I don't know, your notes in another language. It was like really like, you know, and you could see how you got there because we learned about all these pain points, all these use cases, and then what we did, and I think this is probably, probably one of the things I'm proudest of because it was hard, is we, we looked at it all and we cut out 50% of it. We basically redesigned and cut out 50%. And I think that would have been impossible or extremely hard to do if we had been publicly launched. I think because if we had been publicly launched, we had all these people, you know, grown to love granola and whatever weird shape it had been in, and then we cut out half the functionality. You just get, you get so much hate. It'd be tough, but because we were still pre launch, we only pissed off 150 people instead of, instead of the number of people who use granola. Now, simplicity is really hard. And it's hard because organizationally, unless you're, you know, the founder, like you're solving a problem and you're in a little universe and you're gonna, you're going to design for something that's going to optimize to solve the problem that you're fixing, whatever that is, right? But you're not, you don't have the full context of the product in mind, you don't have the full context of the strategy. And you end up with lots of different people going for whatever is a local maximum solution based on the worldview that they have, this is the problem they're solving. And then you have to have this other layer which is looking at the product end to end and saying, sure, there's very clear, tangible value in having this feature. And then there's this like, very untangible, hard to measure cost to launching it, right? And every on a one by one basis, it always looks like you should launch the feature, right? But it's. And then you look up and you have 10 buttons that are clogging up the app. And then, and the app no longer feels so magical and so zen. And the real danger there is user requests like, you know, people always ask for the things they don't have. People Rarely say, oh, actually, can you, can you cut out half of the functionality of the app? Even though when people, when people talk about granola, what they love about it is that it's simple. So basically the only person, the only people in the, in the universe who are going to be pushing for simplicity are going to be the kind of like the designer product leaders in the org. And it's kind of a lonely job, right, because you kind of make everyone angry or everyone's unhappy with you, but it's such an important job.

And to the tangible versus untangible point, had you decide effectively to play back some of what you just said, the 50% that you need to cut. Are you looking for qualitative feedback or do you look at quantitatively what people actually do with the product? Which part is science, art, Taste versus data measurement?

Yeah, it's all of the above. So our, our general philosophy that's gotten us here, and it may not get us there as we scale is we make most product and design intuition, sorry, decisions based on intuition. Like, what do we think makes sense? It's kind of this vision of the product and that we're headed towards and that, like, we just kind of make the decisions based on, like, does this feel right? Does it feel like it's in line with the vision? And what we do to make sure we're not divorced from reality is think of us as like an LLM. It's like we try to fill our context with as much real user feedback, user opinion as possible. And some of that is quantitative, of course. Everything we launch, we measure and we look at the graphs and it's like, oh, people. A lot of people were asking for this and not that many people use it. It's like, oh, it must have been the loud minority. And, you know, that's a very important tool in the tool chest. But what I think is even more important is constantly talking to people. And that's something where it's like Sam and I, and I'm talking about Sam and I because we work very closely together. But like, most of the team actually, they, they do regular user calls. And we, like, we aim to do. I think Sam and I aim to do four to six calls a week with users, but constantly. Not like, oh, we're, we're doing a sprint on this feature. It's actually, we try to book them every day, always, so that there's this constant context of. And the thing is, when you're building product, it's so easy to abstract away a human. So, so, so Easy. It's like it's, it's, it's what our brains always push us to do. And when you abstract away the user, it becomes very easy to convince yourself that they want X or they're going to do. Of course, if we make build this feature, of course they're going to use it. And I think it's only when you have constant contact with people and you're like, oh, yeah, they're so busy, they have all these other things to worry about. They don't even know about what buttons are in granola or not. Like, of course they're not even going to notice that. It's like that kind of thing that's really important to do qualitatively.

How do you think about all of this going forward in a context where presumably you're getting pulled in different directions? So one direction presumably, is the fact that you've been very successful with the tech Silicon Valley crash. But then you're going to go into lots of different industries with people with different experiences, different needs, and perhaps different levels of expectation or comfort with technology and AI. So on the one hand, and then on the other hand, you just launched granola teams a couple of months ago and that pulls you into the world of Enterprise and SoC2 and Compliance and all those things. So I'm curious about how you think about balancing all of this versus simplicity and then how you prioritize the roadmap, you know, with all that in mind.

It's absolutely true. We are, we're being pulled in a million different directions and it's very challenging. I think the overarching point that I have here is, I think the failure mode is that we optimize for today's world. So we optimize for today's product and today's world and today's needs, Needs. And it's easy when you just talk to users and you get these requirements or you talk to enterprises, you just kind of. It's easy to make this assumption that's like, oh yeah, I'll come up with a plan, assuming the world stays static and we are in one of the fastest moving moments in tech history right now. So I think the main failure mode for Granola is not to invest enough in building for the world of tomorrow. And maybe you can kind of infer some of this from what I said earlier about Granola is not about meeting notes, it's actually a tool for thought to help you do work. We're very, very excited about a world where people use granola not just to take Notes but to do all kinds of work. In a way, the product we have today is a Trojan horse to collect a lot of your context so that you can then use all the information in that context to do future work. But that is hard because you have users or companies asking you for feature X today, right. And we have to simultaneously invest in like oh, we're doing a kind of a, this really incredible like deep research mode across that can look at thousands of meetings in a matter of seconds and pull out these insights. And you know that's, that's not something that our enterprise customers are asking for right now because they're not even thinking to ask for that. But I guarantee you that will be a huge part of, of where the world is going and where a lot of value of granola will come from.

I'd love to spend a little bit of time now on how it all works behind the scenes, the tech stack, the mechanics of it all. So starting with the model so seen on the website and as a user you use lots of different models. So as a first question, do you exclusively at this point use third party models? So have you built some of your stuff from a pure AI perspective?

No. Our philosophy is to use the best model that is on the market as quickly as possible. There's so much value in focusing on the product like low hanging fruit when you focus on the user experience today. And the base models are getting so much better and so smarter so quickly. Our strategy has been to use the latest and greatest and when we feel like we hit a wall and the only way to make the experience better is then to fine tune or train models and we will do that. What we found is that there's so much alpha in the improving models and making sure you get the most out of that. That that's kept us busy thus far.

And so that multimodal world, that's OpenAI anthropic Google, is that right? Are there others open source? Any specific models that you currently use that you just announced that you are supporting?

Yeah, yeah. I mean we basically test out all of the models that come out and anyone who spent a lot of time with models, even users are getting really sophisticated. You just learn their abilities. It's like, oh, it's like this model's really good at writing these types of things. This model is really good. When you stick a ton of information in the context window, it can pull out the right stuff. So well, we include all these models in granola. We set the defaults, we set different default models for the Specific thing you're trying to do in the app.

Interesting. So you provide the ability to pick a model, but you gently guide the user towards what's best for their use case. Okay. How do you think about keeping a consistent user experience in a world precisely where those models, one, evolve all the time, two, behave differently, and three, as we all know, are stochastic, not deterministic. Do you expect the users, as you just said, to be smart about it and that's sort of theirs to figure out, or is that something that you abstract away for them?

We abstracted away from people, the general design. At first we didn't. For the longest time, we didn't let users choose their model, and we only let users choose their model on chat. On the note generation side, we completely abstracted away. And the reason we do that is every time a new model comes out, we have to completely change or tweak the prompts that we use for note generation to provide consistency of experience and an improvement of experience. And there's significant work that goes into that. I think it's one of the value adds that Granola brings, as opposed to just working with base models, is that we take care of that. We make sure you get granola feeling or sounding notes consistently and that they keep getting better over time.

And how does the prompting work behind the scenes?

Granola takes in a bunch of signals about you. So who you are, what kind of work you do, where you work, who you're meeting with, where do they work, what are they trying to do? And we've put a ton of work into what are common meeting types with folks with different jobs. And in those meetings, what are the things that really matter? So, for example, I think the thing that kind of blew people's mind when granola came out was if, let's say a vc like an investor and a founder were both using granola in the same pitch meeting, let's say the notes that Granola would generate for each of them look completely different. Right. And as something as basic as I mostly care about what you said in the meeting, not what I said, sometimes I care a little bit about what I said, but it's usually what the other person says. But also the kinds of things that I care about coming out of a pitch meeting is very different as a founder, as it would be as an investor. And a lot of that is kind of hard coded instructions that we build into the system.

How do you navigate the context, window constraints? If you have a one hour meeting, that's a lot of information and I'm curious about you think about chunking.

Sure. So I mean one hour meeting was a lot of information and in 2023 it's not a lot of information now compared to what the models can do. They have incredibly the context window size increase over the last three years has just been mind blowing and fantastic for us.

Yeah, but call it a board meeting which is four hours. Right. And there's like rate limits and all the things. Or maybe, maybe it's no longer a problem at all.

No, no, no, that's not a problem. The problem becomes now. Well, okay, there are two things. One, notes aren't very. Notes are short, transcripts are long. Single meeting is fine. It's when you have lots of meetings and a large corpus information that that that's where this problem comes up a lot. And the like the interesting the trade off or the thing that's tricky here is that if you care about information lookup right then, then you can do rag. You can do well either like some form of keyword search or cosine similarity. What we found is that a lot of the most interesting queries that people.

Have.

Like would completely fail with that type of method. So for example query might be like what are all the things I didn't do a good job explaining? Or tell me what are all the bugs that this user encountered in this user call. The only way you can get a good answer to that is if the model has the full context. This is very costly but we generally tend to put lots of context into the context windows and we err on that side and we have some really cool stuff in the works where we look at full context across thousands of meetings which I've told you about. But again it is costly. With today's technology there are trade offs and the trade offs are basically as far as we've seen it, like money or quality. And our philosophy since the beginning of granola is always to build for the world a year from now because by the time we build it and it gets distribution, the costs of those models or those capabilities will come down to a reasonable place. But that's basically I think the trade off there is around the quality of the queries that need a lot of intelligence.

Yeah. And the cost question is particularly interesting in timely these days because of the well reported discussion around the cursor of the world or the AI coding tools having negative gross margins. Is that a situation where like directionally you guys are at where you for now operate on lower gross margins? And I will not ask you for any Specific numbers. But again, you're building for the world of tomorrow where you have higher gross margins. Is the world of like meetings different in terms of like token needs versus AI coding.

So the most expensive thing about our business is actually transcription and, and historically it's been actually transcription and high quality transcription versus LLM inference. And the. We basically use the best transcription, real time transcription on the market at any time. And the cost of transcription has fallen dramatically over the last couple of years and I suspect we'll continue to do so. So yeah, we're, we're not at negative gross margins right now, but what I do expect is I expect the cost of inference to stay the same or go up. As we allow users to do much more complicated queries over much larger data sets. It'll be an interesting race to see does the cost of inference go down faster than the user? Desire for more complicated, more intelligent features goes up.

Talk about, if you will, the parts around the model. So you mentioned transcription, obviously there is a big audio part to what you do with. There's a lot of problems in that world. There's the problem of diarization, which means figuring out this is Chris speaking or Matt speaking. And there is the problem of noise cancellation. What work have you guys done? Which vendors and solutions have you picked? What have you learned?

So let's see. Echo cancellation, we run ourselves on device and it's just important because if someone has headphones on and they take off the headphones halfway through the meeting, it's important for that to work.

And something you built internally?

Yeah, on top of some open source frameworks. And then we built it. We've partnered with Deepgram and assembly for transcription. They keep pumping out better and better models. We're always using the latest and greatest. What else are we doing? Diarization, unfortunately, real time diarization is still in its infancy in terms of quality. So that's something that we're keeping a very close eye on. But we haven't been able to get real time diarization at a quality point where we're happy with. And actually there's a danger with if you give. Models are really smart in ways you don't expect. If you give like incorrect diarization to a model, it'll oftentimes confuse it more than if it just has to try to infer who's speaking. So there's some interesting analysis and evals to be done there. Let me see. Yeah, we use, I mean there's so many. I can tell you about our whole tech stack. We use brain trust for a Lot.

Of the evals and how do you think about guardrails? Make sure that the system doesn't spit out, you know, things it shouldn't. For example, I guess for every product.

The idea of like, what should the system not what would be harmful or negative for the system to spit out? A little bit different. I think if you go to, if you go to something like Google or ChatGPT and you ask it for something, it's like an open ended place where you're, you're looking for guidance or help or health advice or what have you. There's some really, really bad scenarios there. For our case, it's a little bit different. You're usually going back and asking questions over your meeting data. So there, there's a question of we get something wrong or it hallucinates. But what we found there is that the best thing to do, you're never going to get it 100% right. We can never be like, oh, you know, we make no mistakes, you know, you just trust us. So obviously we do the best we can to avoid those mistakes. But really what's important is the way you design your product needs to let the user kind of like view source, kind of look behind the curtain and be like, how did you construct this answer? What are all the citations? So we spent a lot of time thinking about citations, about letting you view original transcripts and quotes and there's a lot more we want to do there. But that's really been the, the way you solve for this, at least so far.

Switching tax a little bit away from the tech stack. I'd love to go into growth mechanics and what you've learned. So you said, among the many interesting things you said earlier, you said one on that topic that caught my attention. Not having the bot first experience was actually a trade off in terms of virality because you don't have the built in product exposure because there's no bot showing up. So what have you done to sort of overcome that and what are the, you know, viral sort of growth mechanics.

Built into the product today we haven't focused on. Basically what we really focus on is making the product really good for people. And turns out that that's actually led to a lot of viral growth. But that viral growth is from people telling each other like so like an interesting story that I never imagined that this could have happened, but I hear a lot now is if you have a one on like, you know, you're meeting with someone on a zoom call and your AI bot shows up and you basically are Told like, hey, what are you doing with an AI bot? Like, why aren't you on granola yet? You know? And it's like, oh, wow, the AI bot is now like, conversation starter. Yes. The weird thing is it's a conversation starter for a human to bring up granola and to bout for it, which is incredible. I never would have sat down and imagined that world. So we always start from a value standpoint, what is valuable to the users. We could email your notes to everybody in the meeting, like all the other companies do. But again, is that a tool that you want to use? Is that acting like a tool for you, or is that acting like a growth engine? What we do have is we do let people share granola notes. Basically, you can share notes on a link and you can send that link to people. And what's nice about that is that when you share the granola link, the other person can chat with the transcript and ask questions. So it's kind of like unlocking all the AI capabilities but to the person you're sending it to. And we see a lot of link sharing there. And then a lot of times people then say, oh, this thing seems interesting. What's this? And they go and they download granola and they. And they grow that way. The thing we're working on right now, and it's still early days, is if granola acts as a second brain for you. Like, our goal, the next step there is to be kind of a second brain for your team or for your company. And this is how obviously at Granola that we have that. And it is pretty incredible what you can do when you have all that shared context. That it does bring up a lot of questions again of like, what kind of meetings, what kind of context do I want shared with whom, and what kind of meetings or contexts do I not want shared? Because there are a lot of, a lot of really dangerous failure modes here, right? Like, it's very easy to sit down and be like, oh, you know what? I want all my meetings shared. All the meetings in the company should be shared, right? Because, like, wouldn't, like, transparency is a good thing and it's super valuable in the age of AI, right? Like, the more context the better. And then you actually. And then you. And then you actually sit down and think about meetings. You hear these horror stories that go for viral where it's like, you know, like the, the AI meeting notes were like, captured, like a sensitive meeting on the wrong calendar, invite and like emailed the whole company, or, oh, I was at a founder Dinner the other day. Okay, this is another, this is a great story of how, how we get customers. I, I went to a founder dinner and he was like, oh, you're Chris from Granola, my company. We, we just switched to Granola. And I was like, oh great, like what happened? He's like, well, I walked in on my co founder and my CTO having a discussion about letting go of this key person and I noticed that, I think it was like the Google Meet recorder was on and it was on the all hands, which had happened just before in that meeting. And we had this awful realization that the moment we hit end on the Google Meet, everybody in the company was going to get an email with our in depth discussion of how we're going to let this person go. And they, and then they all freaked out and they all said, okay, like if the WI fi cuts out, we're screwed. If the battery dies, we're screwed. So this became like the, the sacred computer. And I think they had like 10 people trying to figure out and they were able to change. They figured out there's like some like undocumented setting in the workspace admin like data control thing and they were able to turn it off, but they like, they turned it off and they ended the meeting and they just sat there for five minutes waiting to see if it was going to be a disaster. But there's a lot like, turns out human relationships are complex and nuanced and if you have a one size fits all solution, there are a lot of these cases that get kind of ugly.

As you get further into the enterprise. Do you get any kind of pushback or questions about what it means for every conversation to be recorded and for it to be a, I mean quite literally like a track record or of everything that I was ever said from a, I don't know, legal perspective or any of those?

Yeah, absolutely. So I think, I think there's two lines of questions here. One is like, it's important in the enterprise context that everyone knows that you're using granola, right? And we have functionality that posts this in the chat right now. You can turn this on and be like, whenever you join a call, like post in the chat, let everyone know I'm using granola and we're going to launch a whole bunch of stuff that makes that better. But that's like one line of questioning and like you should always tell people, use granola. It's like the right thing to do regardless of what the laws of where you are. You know, I think it's like, we're also moving towards a world where like these tools, if they're well designed and not too invasive, like it'll be normalized in certain types of contexts, then the other question is like, and this is not just a question for granola, but this is for AI in general, which is like, there's your liability footprint, right? Like, the more you want to. At Google, our emails were deleted after two years or three years, right? So you literally couldn't go back and search and see why a certain decision was made on a product. You know, like, why did we do this on Gmail 3 years ago? You couldn't find that in the email. And that's to limit the liability footprint. In a world of AI where all that stuff's actually really useful, or we think has the promise of being really useful in the future, there's a lot of tension there, especially when you talk to a lot of customers. The biggest disconnect that I've seen from especially folks who are very AI forward, tech forward, thinking about the future, it's like, how can everyone in the company leverage AI tools to become better, faster, smarter and, and then the IT team, the legal team, and I don't know how that will get resolved. I honestly don't. I think it'll be a very interesting space to watch.

So still, in the, in the growth mechanics world, you have an incredible user retention. Curious if beyond the sheer quality of the product, there is anything that you did and that maybe people can borrow for for their products, leading to success in retention.

From my entire career, like building products, you kind of learn the hard way that getting users to build a habit of using your product is incredibly hard. Like when I was earlier in my career, I thought, okay, you build a fantastic product and then people will use it. And then there's this horrifying moment where you build something good and you put it in front of someone. They say, this is fantastic, right? And then the next week they just hit that exact same pain point. That's the exact same moment. And they don't use your product. And you ask why? And they say, oh, I just forgot about it. I didn't think to use it. No, I still love your product. I just didn't think to use it. And that's a really, it's really heart wrenching to realize that. So now, especially when people, let's say founders, come up to me with their idea, I urge them to think really, really hard about what are the, the triggers for using your product and those moments and the beautiful thing about meetings is that there's a, you know, meetings are on a calendar and there's a very specific moment where we know like you're going to do a meeting and that's not enough. Right. If granola weren't actually useful then, you know, that wouldn't lead to retention. But it's, I think the combination that granola is useful to folks and we can send notifications at the right moment to start it.

All right, so as we get towards the end of this conversation, I'd be remiss not to ask the obvious question about the competitive landscape. So we started the conversation talking about the existing note takers. The sort of elephant in the room kind of question that I'm sure you get all the time is why wouldn't OpenAI do that? And I'm a VC, so I'm a specialist of asking why Google wouldn't do that and why would nobody do that? But zoom, as, as a product and it seems to be so fundamentally important what you're doing and so horizontal and so transformative that it feels like all those great companies should focus on this at one point or another. How do you think about navigating that tension?

Yeah, well, I guess it was really helpful that most of our competitors had some kind of AI note taking feature, but when we launched, so that was already the case and somehow granola was able to stand out and win people's hearts and grow. I think again, the failure mode here is to think about the world as it is today and the product capabilities as it is today. And I think my view is that notes are kind of useful. I think they are a stepping stone to the way we're going to work in the future. And the way we're going to work in the future is with AI that has really deep personal context about you and the product experience that Granola will be a year from now, two years from now will look radically different from what it is today. Hopefully it'll still be very simple, but it'll help you do a lot of work. And I think no one's built that yet. A lot of people are racing towards that. AI is an incredibly competitive space. But when we're talking about, really when you're talking about products that are native to a new medium, oftentimes startups have.

An advantage, but just to press a little bit. I'm mostly thinking about OpenAI because I do think of Granola and you don't need another person to tell you that you're the one building it. But I do think of Granola as building memory for the world and for all of us. And so that is incredibly strategic for OpenAI or Google because that's the ultimate price, right? Like you're building, as you said, like second brain. So how do you think about those players?

Yeah, I, I would put Google and OpenAI in different buckets. I, I would put. So I like, I think a lot more about OpenAI and anthropic than any of the, the legacy players because I think they are native, they are AI native, they're, they've led the way in this space to me, it all kind of. And again, I, I don't have a crystal ball. I don't know what the future is going to look like, but OpenAI is going to try to do everything to everyone, really. And they do an amazing job at it. It's really, really incredible. And I think the question is, can we do something way better for a specific use case and a specific type of user? And I think there are, it's hard to visualize that world, right? There's a world where like, oh, actually, you know, it's not that specific. The upside's not that great. Or actually, you know, what power tools for this, like specific, these specific workflows and just nailing them because you care more matters a lot. And I wouldn't have made my bet if I didn't believe that the quality of the experience and the tailoring for our users isn't going to win. But again, I don't have a crystal ball. I think kind of like it'll be fun to watch. I have high hopes for a lot of the stuff that we haven't launched yet and how different that's going to feel.

So a little bit on that note, to close and zoom out anything that you can do, talk about in terms of the roadmap or the future. So you mentioned a couple of times the idea of searching through the history of meetings. So that's one thing. Maybe double click on that and. Or anything else that you can talk about.

Yeah, absolutely. So I think the world we're moving towards is you have a bucket of context and then you, you generate documents or artifacts on a per need basis on the fly. And the types of things that we're working on are. Given my entire history of meetings, can you pull out really? So for example, a good question you could ask, could be that we could ask would be like, okay, out of everyone we've met in the last two years, who are the firms who are most likely candidates to lead our series? C Right. That's a. A question that I can't ask anywhere else in the world right now, but I have a version of Granola that will go through my 2500 meetings and spit out a remarkably intelligent answer to that in 20 seconds. It's like a deep research mode. Right. The other thing that we've played around with is, like, if you're. If you're dynamically generating artifacts or UIs on the fly, can those be shared? Right. So this idea of we have a. A folder where all our sales calls get put into and they're shared within the company. And then we have this artifact that you go to the URL, it doesn't have to be in the Granola app. And it'll tell you, here are the most important things our enterprise customers are telling us, like today. And every time you reload it, it's up to date, but it's like a. It's. It's like a memo. So there's all these. And there's some use cases I can't. I can't talk about just yet, but basically, like, manipulating information for you on the fly based on your context and increasingly large context, I think will unlock all kinds of use cases and workflows that people aren't even thinking about right now.

And I don't know if that's part of the roadmap or whether you can do this to some extent in the product today, but for me, clearly, where this could be going and where as a user I would find that absolutely fascinating is Granola as a coach, where, you know, based on, you know, everything I do all day, it could be. Matt, you need to ask better questions. Those are the three questions you never ask. You need to ask, or, you know, you spend a lot of your time on stuff that doesn't really move the needle kind of stuff. Not that I do that just as a, you know, as a. Yeah, yeah, yeah.

I'm curious how to. For you personally, who would be. Is there, like, a human who would be the ideal coach? Like, you know, if we're gonna go train a granola coach on a person, is there a specific human in mind that you'd like?

Yeah, and look, I'm not. I'm not very qualified because I never had one, but I think the entire coaching industry, right. Of, like, people that you spend, you know, one hour a week and you talk about how you use your time and what issues you encounter at work, where you could. Where you do well, where you don't do so well, and you have this kind of, you know, sit down session. A little bit like you would get a psychiatrist or a therapist of some sort. If Granola just knows everything I say in all meetings after a certain time, I think Granola is going to have a very good idea of where, you know, I succeed and where I could get better.

Stay tuned. Okay.

Exciting. Very, very good. Well, it's been a fantastic conversation. Again, you know, I'll end where I started, which is I'm a huge fan of the product and it was life changing for me, I have to say as well. At a personal level, you know, when we talk about feeling how AI is both enhancing us but also sort of disrupting us. In my early use of Granola, there was still, it was still just English and you guys, I think recently, or at least I saw it recently, introduced a multi language feature. And as a European French born person who has spent most of his life in the US at this stage, my secret kind of little trick ability that I had was always to listen to a conversation in French and take notes in English directly in real time. And I was very proud of that. And it took me, you know, a lifetime to achieve the level of fluency I'm able to do it. And then one day that feature appeared on Granola and I was, there you go. The machine does something much better than took me, you know, a couple of decades to figure out. So I don't know if that's exciting or terrifying, probably a combination of both. But in the meantime, I really enjoy the feature and the product and I very much enjoy this conversation.

Thank you. I guess the hope is you make better investments now post Granola. I think that's the ultimate question, right?

Absolutely. Hence the coaching part. Very excited for what you've built and for the future of the company. Thank you so much for spending time with us. Plenty of lessons for builders around product and growth and building on top of AI. So really appreciate it.

Thank you. Thank you so much, Matt.

Hi, it's Matt Turk again. Thanks for listening to this episode of the MAD podcast. If you enjoyed it, we'd be very grateful if you would consider subscribing, if you haven't already, or leaving a positive review or comment on whichever platform you're watching this or listening to this episode from. This really helps us build a podcast and get great guests. Thanks and see you at the next epis.