How Gamma Grew to $50M in ARR in 2 Years

We started to build conviction that there was something here in AI. And we basically had a bet the company moment. We said we are going to put all of our energy into one big launch around AI. We're going to do it in early 2023. And if it works, great. If it doesn't work, we're probably shutting down shop. The thing that I didn't anticipate was that there was this viral effect where we would suddenly now be at 4,000 and 5,000 signups a day. And I started projecting the line forward and seeing that if the trend continued, by 3 months later, we would be at like 50,000 signups a day.

More than half of the messages we were getting were people saying, "How do I pay? I want more AI." So, not launched with monetization. We had capped the amount of AI you could use with like a credits model, but we didn't give you a way to buy more because we were just in such a rush to get this thing out. And everybody was saying like, "Take my money." Basically, welcome to Subversive, a podcast dedicated to sharing stories from the best consumer subscription apps in the world. We'll bring you lessons for how to grow your consumer subscription business, including insights and inflection points that led to exponential growth from leaders at category defining companies and innovative startups. Let's get into the show. My guest today is John Nona, the co-founder and chief product officer of Gamma. In the last 2 years, Gamma has become the most popular AI powered presentation tool in the world, having

already grown to 50 million registered users and $50 million in annual recurring revenue. In this episode, I talked to John about how Gamma found initial product market fit, about the viral user acquisition loops that have fueled its exponential growth, and about how it has adapted its monetization strategy to cover the cost of AI powered features while still delighting its subscribers. All right, welcome John to the Subursa podcast. super excited to have you on today to talk about the origin story behind Gamma, which is one of the fastest growing AI companies in the world right now, particularly in the world of AI powered presentations. We'll talk a little bit about your background as well as those early days finding product market fit and then how you've managed to accelerate the growth of Gamma um over these last couple years.

But before we do that, would love to just get a little bit of background on you, John. So, where did you start your career and how did you find your way to Gamma eventually? Thanks, Phil. Great to be here with you. Excited to chat more about it. I started my career in product management. So I've been a product manager for pretty much my entire career. I started at the very big company end of the spectrum at Microsoft. So my first role was as a product manager on Bing, which is actually a neat place to cut my teeth because even though uh Bing was not famous for being the best search engine, it was actually one of the more innovative parts of Microsoft at the time and one that was really building consumer online services at very large scale. So hundreds of millions of users doing very heavy AB testing and experimentation. and I worked in image search which is a neat area of sort of visual AI uh within Microsoft. So got my

start there. Uh learned a lot but also realized that the big company thing really wasn't for me. Wanted to go somewhere smaller. Want to go to Silicon Valley. So found myself at Optimizely a few years later. Optimizely slogan was AB testing you'll actually use. So got to bring my sort of experimentation lessons from Microsoft to a much smaller startup environment. I joined Optimizely after the series A. So it was around 100 people very quickly became almost 500 people and kind of rose through the ranks and also just saw the chaos of building a company at that kind of like series A through D scale which was really interesting. So uh learned a lot there actually met my co-founders there as well as a lot of our early team for Gamma. Uh and in 2020 kind of made the leap to start a new company and that's what led me to Gamma. Very cool. So you started at a large company you decided that wasn't for you. You went to a

smaller company in Optimizely, but then it grew so fast that you were back at a relatively certainly not Microsoft scale. Got the true small company experience. Yeah. Yeah. Well, I guess you're getting that now. And so maybe

that's a good segue into Gamma. So it's 2020, right around the time the pandemic first started. Um, walk me through those early days of Gamma. like where did the idea come from to start an AI powered or at the time maybe not so AI powered presentation uh company especially in a world where you already have uh you know Microsoft and Google Slides or or was it even a presentation uh company when you first started it? Well, first of all, you're right about the co timing being very relevant. Uh, I actually remember uh meeting up with my eventual co-founder Grant in London in I think it was the first week of March 2020. Uh, right before the world ended basically, but we didn't know that yet. We had just a a great dinner in London. That's where he kind of first floated this idea and we started chatting about it. Um, little did I know it was the last time I would see him in person for like two years um

because of everything that was about to uh go down. But yeah, the original concept for the company started a few months after that when we were already in the depths of the pandemic and we just both had this realization like uh you know, we've been working on we all of our job all day is just making presentations and sitting through presentations but then we're like watching them on our phones from like the park because we're all remote working or you know from like the desk I tried to create out of my bathroom because there were no rooms left in my house. So there's this feeling that like the idea of presentations was obsolete.

It was made for this different earlier world and the world was changing fast and so remote presentations was the original concept and also asynchronous presentations. So creating a new format or medium that was actually somewhere in between presentation and document something you can send around ahead of the meeting or consume after the meeting without actually having to be there. Our original slogan uh I think was uh something like uh presentations you'll actually want to sit through or something which didn't quite stick. We went through a couple versions of that, but that was the original mindset. There was no AI in the picture at all. It was really all about future of work. That was sort of the big investment thesis that we wrote on, and we actually rose raised a precede off of that kind of like concept. We built quite a lot there. But fast forward a bit to 2022, we had built a beta version of our

product. Um, we'd seen some traction, but certainly not exponential growth. we were in kind of that uncanny valley of some usage, some happy customers, but not enough to justify another fundraising round and not enough to really be sure there was really something here. Um, and I think many companies got caught in this crunch of like the very high expectations of 2020 and then the reality of the starting to crash and also just things going back to normal in 2022. Maybe maybe the future of work is still the present of work.

So, we kind of got to this point of having to rethink our core thesis for the company. We thought we got a lot of things right about the product. Uh we thought we were really right about this idea of merging presentations and documents and having this more fluid, more modern medium that wasn't so rigid. We were sitting there having launched our beta thinking, man, we have this huge blank page problem. Um and and the way we thought of it was we have a blank page problem. We gamma trying to increase our activation rate are seeing that like 90 95% of people sign up for our product, use it for a few minutes and just don't really get it. They bounce and they leave forever. And so the sort of core goal that we set for ourselves was how do we raise that activation rate? How do we increase the percentage of users who hit that aha moment and get the product enough that they'll come back and use it and

eventually tell their friends? And so we thought if we could give you an incredible first fiveinut experience of using the product, you would then stick around yourself and tell others about it. And so AI kind of emerged as possibly the perfect solution for this because instead of your first 5 minutes being here's a blank editor, go build something, the first five minutes could be, hey, we'll start you off with a rough draft of your presentation in this new gamma format and then you can go from there. You can edit it, you can refine it, but now you've sort of got the value. We started to build conviction that there was something here in AI and we basically had a bet the company moment. We said we are going to put all of our energy into one big launch around AI. Uh we're going to do it in early 2023 and if it works great, if it doesn't work, we're probably shutting down shop. And I think I

mentally gave us maybe I don't know 30 or 40% odds. Uh I just thought it sounded a bit preposterous that it would all sort of like come together and work. I I also had this like how could it not work? Like this is so freaking cool. This is such a game changer. You know, nobody's seen something like this. And it all really came together just in time, like literally a matter of weeks for us to do this launch. Uh we finally launched it. And uh funnily enough, actually what happened was in between my two ideas of like it'll blow up like crazy or it'll just fizzle. It actually wasn't clear right after launch if we had made it or not. We we did see a nice big bump of users. I still remember the way we launched was through a tweet.

Actually, we decided to bet it all on a tweet with a big high production launch video. So, we we made nice launch video and then Grant, my co-founder, uh made this tweet and we were in existential like company's going to die mode. So, we're like, what is the most provocative clickbay tweet we can possibly do that will get people to pay attention to our product and watch this video and sign up? And the thing that we landed on was something like um the most important task in business is about to be obsolete. introducing like AI that makes your presentations for you. And that actually worked really well. The tweet did go viral. Not immediately. First day it went out, we got like maybe like hundreds of likes and some comments and we got some signups. But then like 3 days after we launched, um I think Paul Graham retweeted it and a couple other really big VCs retweeted it. Yeah. But but but the thing that he seized on was not,

"Wow, what a cool demo." The thing he retweeted was this is so stupid. Um, presentations are not the most important part of business. If you think this way, like you're not a good founder. But it didn't matter because he did this and then several other VCs piled on and replied that the provocative tweet was incorrect and we just got tons of views, tons of eyeballs and it started to pick up and we got more and more signups. I want to I want to jump on one thing you said there because I, you know, I've spent time as the first employee at a com at a company. Uh, I founded my own startup at one point. I've spent a few years in venture capital and and a consistent pattern I've seen especially when there are these new technology waves like mobile or now AI oftent times it's the nonobvious you know not necessarily 10x better but very different startup ideas that win simply because they're different enough that they spark a

conversation and often times it's a very polarizing provocative conversation so in this case you have Paul Graham on one side saying this doesn't make sense this isn't the right place to start but then you have other people jumping in and saying no actually this is a really important use case. So, I imagine it was difficult in this moment, this bet the company moment where you have this big launch, you have this viral tweet, but there's still sort of some mixed signals. What gave you and Grant and the rest of your team confidence to move forward after that launch? Was there a particular metric or particular set of feedback you were getting from customers that told you, "Oh, we're on to something here." In the immediate aftermath, we did see a spike in signups, but it started to subside. And so that's where we thought maybe this thing isn't quite gonna cut it. But uh then what changed is we started to

actually see even after an initial trough of new user signups, it started to come back up again. So we started to see just a few more signups every single day. It was like 1,000 and then 1,500 and then 2,000 and then 2500. Ironically, the metric that I thought we would fix with this whole launch, which was our activation rate, did not budge. It was still like stubbornly low. uh by which the launch was a failure. But the thing that I didn't anticipate was that there was this viral effect where even the 10% or whatever who did activate were telling people about it because we were seeing more signups every day. And what really stood out to me was not only was it increasing every day, but it was increasing with a surprising regularity where we would suddenly now be at 4,000 and 5,000 signups a day. And I started projecting the line forward and seeing that if the trend continued by 3 months later we would be at like 50,000 signups

a day. And I was like, okay, but there's no way that's that's not really going to happen. But uh it just seeing that growth of new people coming in was proof enough that like, okay, there's something here. There's something self- sustaining. Yeah. And we were also getting a a flood of feedback from these users. They were all using it and telling us everything that was wrong with it. Um, I actually had an automation at the time where every new user who signed up would get an email uh coming from me. It looked like I had written it by hand that said, "Hey, I'm John. I'm the co-founder at Gamma. Just wanted to see if you had any feedback for us based on your experience of using the product." And this had been going for I think a whole year before we'd done this AI launch. And I would get like one or two emails uh a week and they'd be kind of interesting. Suddenly, I was getting 30 or 40 responses a day

to my email and then 50 to 100. And actually it became a large part of my day just responding to all these things. But suddenly just the level of engagement and like the way people were sort of like expressing their frustration was totally different than before. They actually cared. Yeah. Well and often times when you get frustrated feedback like that, it's actually a positive signal because if somebody cares enough to give you constructive criticism, it means they're using your product. They like it enough to bother to give you that feedback, but there are still gaps and so you want to fill those gaps. I'm curious. I think this is a great lesson for other founders out there who have just gotten this flywheel going where their customers care enough

to provide them with this input, but you need to be able to separate the signal from the noise. So, as you started to get dozens and then hundreds of points of feedback per day, how did you filter it down to the most important things that you and the product team then went to work on in terms of building the next most important feature? Well, I think one thing was I actually missed the point that you made earlier at the time, which was I I didn't quite realize that frustration was a positive sign. And so I think the thing that I wish I paid more attention to was what did the frustrated people have in common? Like what were they actually trying to achieve? I wish I more time I kind of listened to the feature requests, but I wish I'd spent more time saying like what is the suddenly like critical work you're trying to get done with this thing that these feature requests are getting in the way of. And it turns out

that there were like these patterns already emerging that like we were starting to see some of our ICPS emerge even then although I didn't realize it yet. So one example of that was we started seeing all of these consultants who were like using our product to make you know like uh they are like sales proposals or like pitch information for clients. There was this pattern that emerged very early on that our product actually had traction for external usage. So people using the product to pitch outside their company to sell something whereas I coming from a career in product management had always thought that our product would be used for internal usage like a company all hands or something like that. And so I think I indexed on the feedback that was most familiar to me. I I paid attention to the people who were PMs like me telling me all the reasons they could never use this for their internal presentations. I

didn't pay enough attention actually to the consultants who were nothing like me talking about how they were already using it to power their business which of course is the real signal. Well and there to me there are two lessons in there. One is don't get lost in the averages. Make sure that you are segmenting your user base. And you don't need necessarily need to have a segment for everyone, but you need to know who your maybe three to five most important customer segments are, especially with a proumer SAS tool where you're going to have different types of people using this product. And then the second one is just make sure you aren't overly biased in terms of overindexing on the feedback that resonates with you simply because of your own background or or set of

experiences. So you launch this new AI powered version of the product in early 2023. great timing because chat GBT has just exploded onto the scene and so people are trying AI products left and right both because they're new and exciting also probably because there's a little bit of fear of missing out or being left behind. Uh but then at some point I I imagine you needed to make sure you really did have strong product market fit. Um and so what were some of the features you then worked on after that early 2023 launch to sort of get that activation rate up and get that long-term subscriber retention up? And what was the point at which you knew that you you had that um stable long-term subscriber retention you you needed to really start to stack stack up more and more and more subscribers, not just visitors who were sort of touring the product experience. We were even more premature than that. We weren't

even monetizing yet at this point. So this was actually the first sign of product market fit, I would say. And I'd always heard people say, "It's hard to define product market fit. You'll just know when you have it." Which was a very frustrating definition. Um but it it really happened to me in this case which was for the first maybe two or 3 weeks after our launch I was still kind of unsure do we have it or do we not and then the thing that started to break one definition of product market fit is that things start to break like break because you have too much usage as opposed to you putting all your time into getting usage. The first thing that broke was actually our support channel. So I mentioned I was sending all these feedback emails. We also had intercom support. What broke is our intercom went from like three messages a day to 50 messages a day to 300 messages a day in a very short period of time. And we

didn't have a support person because it didn't make sense to hire a support person when we had three messages a day. I was the support person, me and some other people on the team. We had like a meeting where we would review the support messages every time and it just got so out of control from all these people that we were spending all of our time not building product, just responding to support messages. That is an excellent signal of product market fit. Another excellent signal of product market fit was more than half of the messages we were getting were people saying, "How do I pay? I want more AI not with monetization." We had capped the amount of AI you could use with like a credits model, but we didn't give you a way to buy more because we were just in such a rush to get this thing out and everybody was saying like, "Take my money basically." And so that was the very first product thing that we

prioritized quite rapidly which is like oh my gosh, we need a Stripe integration. like we need a way for you to pay to get more credits or unlimited AI. And that was the kind of thing we were racing to build just like deflecting support issues, creating this monetization channel. And a few months early, we were at the point of like thinking we have to shut down the company because we need to raise money. But almost immediately after we built monetization and started charging people, we suddenly had this tremendous amount of cash flow coming in. Uh more than we'd ever really expected. And to give you a sense of this, uh, in May of that year, we launched monetization. And by August, we were cash flow positive.

So, we had like more money coming in the bank than was going out. And suddenly, it's like, wow, uh, all these runway problems are just in the rearview mirror. We don't even need to think about them anymore. We can like go focus on other things. Yeah. What a good what a good um turn of events. And I guess another follow-up question that comes to mind given I've been hearing this from more and more AI companies is you know historically with consumer or proumer SAS businesses oftent times the marginal cost of service subscriber was next to zero. Um but now with these AI powered companies there are non-trivial costs associated with delivering the AI powered features. Those costs are coming down rapidly but they still exist and I imagine in 2023 they they were much

higher than they were now. So how did you think about your early monetization strategy and things like subscription pricing and packaging in the context of a product that had non-trivial underlying costs? We sort of approached it from two angles. The first was that we had to set pricing in a way that sort of was commensurate to that cost. And I had initially thought we would price our product the way most proumer productivity tools are priced. So think of like your notion, air table, uh slack type tools. And there was really a standard of $10 per user per month. That was just sort of an established price point. Uh that would be kind of a no-brainer. But uh given my background in experimentation, I decided we would AB test it. Let's just AB test our pricing and try a few different price points. So we tried 8, 10, 12, and 20 bucks a month. And we measured conversion rate and revenue. And I was

shocked to find out that 20 bucks won the test. uh which was the first realization for me that willingness to pay for AI tools looks very different than standard subscription products. I think uh users understand that there's a cost involved which surprised me because this was very early in the AI days. I think it's more standard now. They also seem to uh just get more value from it. It was so easy for them to justify $20 of value in a way that is not true with a nonAI tool. It was just like, "Oh, of course you saved me 30 minutes. that's worth this much of my time. And so we tested our way to it. But the other side was we've actually always been paranoid about costs at Gamma. I think it might stem from the fact that we we built all this AI stuff at a moment of ultra low runway and funding. And so we were paranoid. It's also because we are very latency conscious in our product. So we've always tried to really build

something fast. And that came from this feeling that AI was all about solving this first five minutes problem. you had to have a great experience very very quickly to see value and stick around. And so what that all means is we've actually for a long time this change later really built with the cheapest AI models we could find. So we always said how can we solve this problem with a cheap and fast model not a big expensive slow model. And so that's part of how we got to cash flow positive so quickly.

Even though people thought we were spending millions and millions of dollars on our AI we actually managed to do things fairly cheaply. And we can talk about this later, but uh we actually had to reverse that as we grew further on because we realized that we were actually underserving our customers by kind of underpowering the AI we were giving them. Well, and I imagine part of that, we talked about user segmentation a little bit earlier, but I imagine some of your early adopters, you probably had some of the solarreneurs and the scrappy startup founders who were just looking for every edge they could find. But then as you build a brand and you build a reputation, you start to get people at larger companies who maybe are a little bit more riskaverse. And so you you need to provide a more stable kind of minimum bar of of quality. Um and and also you could afford it because at that point you had a $20 per month subscription and

you were making plenty of money. I was also going to say there's this concept of a Velin good which is the idea that certain products often times they're luxury goods but I actually think some AI products follow the same pattern because if your brand is strong enough then at some point a higher price point can actually signal quality. And in this world where AI products are evolving rapidly and people are trying to choose between uh all of the products out there and find the best ones within reason, a higher price point up up to a certain limit uh can actually signal quality.

And so I I've seen some indication of that with some of the other clients we've worked with as well that it's not necessarily a lower price leads to more conversions. Sometimes a higher price can actually lead to equal or even higher conversion. It's unusual, but you do see it in certain cases and I think new technology waves tend to be one of

those. Um, so let's talk a little bit more around your organic growth loops because while you're doing more and more paid advertising, I think organic growth has been the majority of your growth up until this point. Um, what drove the the original organic word of mouth loop after that um, 2023 launch? And then how have you found ways to lean into that with product features that make it easier or more interesting for people to share their presentations with colleagues or or with other people they know? the original version of our product that had had that big launch.

The only way that you could share it was similar to Google Slides like presenting within the browser. So go you would present and people would see it. You could also send a link to somebody uh if it was more of like a document type format. And uh that went pretty well. And as soon as we launched monetization, we added this made with gamma badge in the bottom corner of everything where paying would remove it. So it doubled as both an upsell lever and a virality driver. Um where people would see this thing Um, even before the badge though, I was surprised to see how much virality there was. And I think what happened was we had put all this time in our years in the wilderness making these really cool looking presentations where it was very easy to make something that looked polished. And so people would just physically ask in person, how did you make that? Like what did you what did you make this with? And people were

excited to share and uh tell each other about it. So that was nice. Um, but we realized that this sharing loop was sort of the core thing that powered our growth. It was also one of the biggest areas of product feedback of like we need more ways to share the content. And so between those two things, it was kind of a no-brainer that sharing had to be a core pillar of the product. So the biggest things that we worked on after this were PDF export and PowerPoint export. So integrating more with the wider world of other tools people work with. And those two things are probably the single two biggest features that unlocked growth for us because they both gave people the trust that they could work in our product uh without being like locked into it. So it kind of got them out of the wall garden. But they also created this whole new virality lever of now all these PDFs and powerpoints that are flirting that said

made with gamma on them and they would just spread around and they exist forever. I mean our data shows now that people are clicking these little badges on PDFs that were made like two years ago and they're still finding our site. Yeah. Well, and I think one important lesson in there is there all different types of virality, right? There's the offline word of mouth where you where you tell your friends, your family, your colleagues. There's online virality where you're posting on social media networks like Tik Tok or Instagram or in your case maybe LinkedIn or X. Um, but then there's also this lesser known form of verality uh that gets referred to as a casual contact loop where it can be so powerful because the activation energy is lower to to actually activate it, right? You're not having to physically tell somebody about a product. They just see the Gamma logo on a presentation.

And especially given how unique and different your presentations look versus PowerPoint or Google Slides, people are naturally curious and so they ask about it. And if you there aren't many products that lend themselves to that type of viral loop, but if you can get it to work, it's super powerful because it's passive. People don't actually have to actively promote your product. It just happens. Uh, and I I had Gary Timman from Grammarly on the podcast a couple months ago, and he he mentioned some of the same things, right? Both offline verality among uh colleagues working at the same company, but then the sort of little Grammarly watermark that hovers over your monitor. And so if somebody is watching you work at your computer, you're giving a presentation, they see that and they naturally ask, "What is, you know, what is that

watermark?" And it leads them in this case to gamma. Um, another organic loop that um, is really powerful if your product is naturally built for it is SEO. And I know this isn't an area that you all have naturally focused a lot on to date, but I'm wondering given all of the userenerated content that gets spun out of Gamma, how are you all thinking about the potential to use SEO as an organic growth loop going forward?

So this became yet another sharing channel we realized was a total another example of surprise product market fit was people were not just making presentations with our product. We'd had this idea that we were this hybrid interactive presentation/document. People started making little websites uh with our product. They would they would start with a pitch deck and then they would realize that their pitch deck was basically just the website for their little company that they'd created and so they started looking for a way to publish those. We wanted to diversify beyond just presentations. We saw this potential to be this much more like omni content platform. And so we built this website publishing flow in our product where you could take what you'd made and publish it either to your own custom domain like philarter.com or you could publish it to philarter.gamma.site for free and we would bundle that in. So we kind of

always looked for what is the free virality version of our product and then the paid premium. So we started building these gamma.site site domains, letting people publish things to them. Um, and those became this natural SEO driver as well, where you would search for any old topic, you would land on a website that was something.gamma.site, it would have that same made with gamma badge in the corner. And so suddenly we had all this SEO accumulating around gamma. That's just on the website side. We're also exploring more this kind of very tried and true method of templates where you can publish pitch deck templates or whatever it is and those all lead back to us. So in general, we're looking to build this viral loop as much as possible and add on these extra layers of it, of which SEO is probably the next big loop we want to invest in. Yeah, that makes a lot of sense and I love how you broke that down as far as you can

get value from users in multiple ways, right? The most obvious way is they pay you for your premium product, but there are also plenty of ways where users can generate free growth for you, either in terms of word of mouth or in this case creating content that then leads users to discover you via a Google search. Um, and so it sounds like you've already started to lean into that with websites. It also seem we can talk a little bit more about this. We you're starting to lean into, as I understand it, using presentation content as potential SEO fodder if a user decides to make a presentation public. And I would imagine that they have the option to toggle a presentation to be private. But just like Quizlet did this with digital flashards, you can do this with slides and that becomes indexed and searchable online. And so is that something you're thinking about to unlock growth both domestically, but then I imagine

particularly in international markets, this is powerful because it gives you kind of a a flywheel you can kickstart right away when you when you move into other countries. It absolutely is. Yeah, it's a huge part of our strategy is basically trying to we've been starting to call it community-led growth. So we have productled growth, but how do we lay on this next piece of all the users in our community lead growth by publishing content online? And part of the insight here was realizing that our customer base was not the base I expected. that there were so many more of these sort of soloreneur external sharing type use cases where often the thing they're using Gamma for is to promote their business and so when we can actually partner with them to for example feature their presentation on our website that feels win to them we get this content asset that drives SEO they get more eyeballs to their business

and everyone feels like they came out great yeah I love that and I think there's a huge amount of opportunity there I mean we saw it with slides share over a decade ago which ultimately got acquired by link LinkedIn and in many ways you all are doing the same thing 10

years later with an AI powered version of the product. Um so I hinted at international expansion there and I know that I believe well over 50% of your usage comes outside of the US. So how have you all thought about product localization? Like to what extent have you invested time and resources into at the very least translating the product into the local language, translating the subscription price into the local currency? But then also, are there any cases where you've actually built distinct product features or landing pages specifically for um a highv value international market? Yeah, the numbers are even more extreme. I would say of our free user base, it's probably 90 95% outside the US and even among paying customers, it's more than 3/4 outside the US. So, we're ultra international.

Uh clear majority of our paying customers don't speak English and so we were probably actually a little slow to adapt to this reality. I think we we actually had a lot of assumptions going and it turned out to be wrong. So we had the assumption that international markets would retain much more poorly than uh US markets. Uh not necessarily true although they did need different pricing. So we have tailored pricing to local currencies. When people were paying in US dollars they retained a lot worse at at those prices. But we could adjust the price point to make sense. We actually have seen in some cases better retention in international markets. The other assumption that we had was that localization would be really really hard and therefore had to be tackled country by country in a fairly grueling way. Um and I think this is another place where AI has really changed the equation because I think I had a lagging

perception from 1015 years ago that localization is really hard. What we learned is that in some ways it can be very easy. The biggest way was that actually what our users craved was localizing the AI output itself. So the model needs to output things in Portuguese or Chinese or whatever it is. That turned out to be literally a oneliner. It's so rare that a major product innovation is just one line of code. But literally just putting one line of code at the end of the prompt that says write in Portuguese just solved localization for us. But from our perspective, the oneliner suddenly meant that people in Brazil could make things in perfect Portuguese. And we're like thrilled with the output. And this, you know, 10xed our market overnight just by supporting this. Other things have been harder like localizing the product. But even there, we've benefited a lot from AI which can do machine translations

that are not perfect but turned out to be good enough to get that first phase of growth. Now that we are much larger, we're starting to do uh human translations where we actually go through in really high quality and localize the UI. But I would say that wasn't a blocker to moving into these countries. It's an optimization to make things work a little better in those countries. Well, and I think so two points here. One is you're right that the amount of time and effort required to do localization well has changed dramatically in the last 10 years and I did a lot of international expansion work at Quizlet that was only 5 years ago and still it was a very different world in terms of the amount of resources required to uh deliver a good internationalized product outside of the US or outside of English speaking markets. Um but the second thing is I think some companies look at international expansion and localization

as a monolithic thing when in fact there are these different stages you can go through starting with the lowest effort stuff and the lowest effort stuff is often just make the product available in English. You will get some early adopters just by doing that. The next step is localize the product Chrome in the local language right that make sure that the UI and the and the core user experience is something that um people using the local language can understand.

The third step is typically the content. So you mentioned that in this case it was very easy for you to change one line of code and now all of the output spit out by um chat gbt or whatever other underlying model a company was using to populate their presentation or website was in the local language. And then the last step and usually the lowest ROI step is building distinct products or features that need to be customized for very specific market use cases. And some products need to do that more than others depending on the nature of their use case. U but if you start with the lowest effort stuff you can you can usually if you have a strong product that people are excited about and it's differentiated enough from competitors you can get a lot of traction just by making it available in English and doing some basic localization before you move on to the really heavy lifting. Year one was only localize the content. Year two

was basically localize the product Chrome. Now we're in year three of this journey and now we are doing that finer grained like local for example we've we launched right to left support of product editing. So that was like a whole feature we had to build. Um we're doing we're starting to do things like localized marketing campaigns at small scale with things like advertising. So now we're starting to be in this more optimization stage. But we did not need to do it all at once. And if we had we probably would have not done it all.

Yeah. which makes sense because you're large enough that there's there's enough value at stake where the ROI is there even for those bigger feature lifts like the right right to left editing. Well, let's move on to monetization. We talked about finding product market fit. We

talked about the organic growth loops that's driven user user growth. Let's talk about monetization. So, you currently have these two subscription plans, Gamma Plus and Gamma Pro. How did you land on the optimal approach to separating features into those two different plans? like who are you trying to serve with each of those distinct premium offerings? It's not clear that we have lay on the optimal approach, but we've got a model that's worked pretty well. We started with only the Pro.

That's where we began at the $20 price point. I I think we then realized there was an opportunity for a lighter plan underneath that at $10. So, uh, we started out being pretty afraid that that would cannibalize too much of the 20. And so, we were conservative. We made a plan that was still pretty limited AI. And when we launched it, I think the split was roughly one/ird of people taking that plus plan and 2/3 taking the pro plan. So we had a fairly topheavy pricing model. Over time, as we have built more and more, we've actually decided to change the optimization in the opposite direction. So we've put more and more value into the plus plan.

Um so that now we're at the point where probably twothirds of people are taking the plus and only one third are taking the pro. And we did that through a whole series of kind of microAB tests of just like making small changes of where features landed, testing the price point, and just looking at how that changed results. The biggest thing that's given us confidence in putting more people into that lower plan is that we've been measuring retention by plan type. And we've actually found somewhat surprisingly that the lower plan type can retain as good or better as the higher plan. And one reason that happens is that we look at net retention. And so when you start low, you have room to go up. But when you up, you don't have as much room. The other one was I think it we realized that we had these two different customer types. Uh one type of customer was more sort of um episodic user who needed AI at certain bursts of

time, wasn't necessarily totally bought in doing everything with the platform. Having a lower price point for them uh was actually uh better because it gave them a easier starting point and if they did churn, it was sort of like less of a hit to us. different type of user was more committed and bought in and wanted full power and clearly cared about professional use. And so we've started to differentiate those more, but I don't think it's totally a solved problem because I think we actually need to make our pro even more differentiated. And we're considering adding a third tier on top as well as we build out more premium features, more business focused features. Yeah, I love the point you made there around net revenue retention.

This is something that a lot of consumer subscription apps don't get to think about as much because they only have a single subscription tier. But with proumer models, it's easier to offer multiple subscription packages just because it's a more highly considered purchase. And so your point around make sure the price is low enough to get users past the initial purchase decision, make sure you're delighting them, maybe overd delivering on value relative to price out of the gate, but then for the people who are really delighted, you can upsell them into the pro plan. I think that's a really compelling point and it's also um very compatible with your organic driven growth strategy because if you're delighting people, if you're delivering more value than you're extracting through price, that means people are going to have high net promoter scores.

They're going to be more likely to tell

other people about your product and so that that drives free acquisition. Um we talked about this a little bit earlier, the fact that because you have these AI powered features, there are underlying costs associated with many of them. I know you have AI credits that you look at in addition to these subscription plans. So, how have you thought about um the balance in terms of making sure that power users who are subscribers, but they're using your product so much that those costs might be racking up over time. Um that that you're you're accounting for that corner case, but then at the same time, you're not making uh things too complicated with both two subscription plans and these AI credits as as sort of a separate currency. Yeah, good question. uh we've optimized probably more for simplicity versus um and that's pushed us the direction of more highle subscription number versus pure usage based pricing. I've noticed

that many other AI tools have gone the opposite direction where they've really emphasized like this usage based pricing at the expense of simplicity and I'm sure it works in their domains but in our world the thing that we're really optimizing for is ubiquity. We want millions of people to be using Gamma and that organic growth loop that you mentioned rather than value extraction. We're not trying to maximize revenue right now ever. Um and so that's kind of informed our approach of like simpler buckets. What that means is that we know we lose money on some of our users. Um and that's okay as long as we're making money off our users on average. So for each of our pricing plans, we actually measure our AI costs very closely uh per individual user and per plan in aggregate. And so we try to make sure that our average cost per plan uh you know is in the black as they say. Uh but if we look at like 90th percentile,

we're totally okay if we're losing money at that point as long as it's not totally outrageous. Yeah. Well, and I think that makes a lot of sense. I think simplicity companies tend to undervalue simplicity and especially when you're dealing with individual consumers and proumers versus enterprise sales, uh where you don't have the benefit of a salesperson to sort of explain all the the bells and whistles. I think simplicity goes a long way. Uh, and then the second thing is if the worst case scenario is that you have a handful of power users who are actually, you know, racking up more cost than they are paying you back in revenue, that just means those users are going to be your biggest promoters and they're going to be telling everybody about how great gamma is. So, it's sort

of a good problem to have. Um, last question on monetization strategy. So you mentioned before that well over 50% um of your usage comes internationally and and that you all have thought uh and this is something we worked on together about international price optimization and making sure that um the price in every given country matches local willingness to pay but that could be complicated in a world where you're operating in dozens or even you know more than 100 countries. So, how have you all thought about again the balancing act between making sure that the price is reasonable in in a given region or country, but uh that you're keeping things simple enough in terms of tiers or or uh buckets of of international markets that it doesn't get too complex. Yeah, we found it helpful to tier countries and just say like we have, you know, a tier one, a tier two, and so we'll do custom pricing for some of these, but we're not custom

pricing for every country on Earth. We just we don't have the bandwidth as a very small team to keep that up. But we found I to give you a sense like we probably have around eight or 10 countries that we have localized pricing for where it's in their own currency at market appropriate rates. Everybody else is still kind of just paying a general dollar rate. And you know, I feel kind of bad because uh there's probably other people in those countries who are missing out on the chance to use Gan because we haven't done that yet. But it all just comes down to bandwidth on our part and our ability to kind of layer these things in at a reasonable pace.

Yeah. And this is what I typically see. Usually it's three tiers or five tiers when you're still relatively early on and then maybe there's a certain set of 5 to 10 countries that are driving the majority of your revenue. There's typically a pretty steep power curve and so for those countries you'll sort of do tailored country bycountry pricing, but for everyone else they fall into one of three to five tiers and that just keeps things relatively simple and easy to manage. So last question before we get to our lightning round. Obviously you guys are one of these hot cutting edge AI companies right now. You're getting all of this viral growth. Um, but AI continues to evolve so rapidly and I

would imagine given your success there are going to be more and more copycats and competitors and so where do you see the world of AI powered presentations and websites going over the next couple years and how do you maintain that advantage? Yeah, I mean we view maintaining that advantage as the number one priority for us you know for us to grow durably. We have to be constantly staying ahead of competitors. Where we see the market evolving here is we think that AI will do more and more but also paradoxically that AI is not the main differentiator. So maybe I can try to explain how both those things can be true. AI doing more and more means that if you think about just all the stages of making a presentation from coming up with a highle story and doing the research to you know like writing the text on each of the slides and making the imagery to then revising it when you get feedback. So like take these 10

slides and make them 20. AI is going to play a role in every one of those steps and it's going to feel like you have a highly paid presentation designer at your beck and call. So you have like whoever makes Steve Jobs keynote like that person works for you now and uh they cost you 20 bucks a month. That's like the world that we're going to and it's going to feel pretty magical for the people making these things. It's going to feel a lot more seamless of just AI proactively making suggestions or getting pulled in reactively. Uh we had a prototype recently I really liked where um if you are if I'm making a presentation to you because you're my boss and I say hey Phil here's the first draft um you could actually take that first draft and leave a comment on it and say I think you should actually split this into two slides and our AI could just do that as the proposal like here's the proposal back to me the maker

and say what do you think John here's what we do with Phil's suggestion like does that look good that sort of interaction will be everywhere the flip side of this is first of all I think you'll stop even noticing that it's AI we're not going going to just have sparkle buttons every time this happened. It's going to just feel like a seamless part of the workflow. Um to the point where any tool that doesn't do this will just feel broken. That's what we're optimizing for. We want PowerPoint to feel broken after you've used Gamma for a bit. But the reason I say that AI is not actually the main differentiator is to make all of this work. There is certainly a degree of like prompt engineering and fine-tuning and all of that. But really, we actually think a lot of the work for ourselves will be in the sort of core product primitives that power all of this. So, how do we have all of the right visual building blocks?

How do we make it so seamless to take what you've made and send it over to these different tools that AI just has an easier set of tools to work with? Yeah. Well, to me, the key meta point you made there is at the end of the day, AI is a means to an end. And right now, AI is so hot and everybody's talking about it. And so the fact that a product is AI powered as a selling point in and of itself, but I imagine that over time that's gonna fade into the background a bit more and ultimately what people are going to care about is the job that your product can do for them and the ways in which that job makes their lives better.

And so whether it's AI or not behind the

scenes doesn't matter as much as the ultimate output that they receive as the customer. Um well this has been great. Let's get to our lightning round. So I'm going to ask you a series of rapidfired questions. I would love just quick one to two cents responses. Uh the first one is a question I always like to ask which is just what is your favorite part about working at Gamma and what's one thing you would change? Uh it's funny because they're uh related. One thing I love about Gamma is we have a very fastm moving in-person sort of demo driven culture. People are just always making stuff throwing it out there getting feedback. You sort of have to be in the office to follow it all and keep up with the buzz. We don't have very tight boundaries between teams of like only you can work in this area and that sort of thing. I love how sort of loosey goosey it is. But it's also driving me crazy as we scale and we're

driving each other crazy with just the sort of frenetic pace. And so we're working on figuring out how do we impose the right layer of structure on top of all of this without breaking the magic that's helped us move fast so far. Yeah, I love that. I mean certainly better to be more action driven and you know demo driven than the alternative where you're just writing strategy decks all day. Um but there's there's a time and a place and so balancing that as a co-founder uh and and chief product officer makes sense. Uh, next question. Outside of Gamma, what's your favorite consumer or proumer subscription app and why? Uh, it is an AI tool. I just love Claude. I like I live my life through Claude at this point. Um, I it's it's my constant AI assistant. Um, I think I only pay like 10 or 20 bucks a month for it, but it's probably worth much more than that to me in terms of the place where I can

just solve such a strange variety of problems. I do all my cooking through Cloud. I just tell it what's in my fridge and I say like, what can I make with this and how do I do that? Um, Claude is fixing my like back pain from traveling. I just like I It's amazing that one subscription can do so many things. Yeah, it really is magical. And I will also say it's funny to hear people talk about, you know, Chachi BT versus Claude versus Perplexity versus Gemini. It reminds me a little bit of the early days of ride sharing where you had Uber and Lyft as the two ultimate leaders and they had such different personalities. So even though they were ultimately offering the same service which in that case was ride sharing in this case you know it's answer engines but they did it in such different ways and they had such different brands and personalities and and sort of core values and ethoses and that really bore

out in the product experience and I think to a degree that's happening now with the LLMs. Um so it's fascinating to watch and and cool to know that you're a Claude guy. Uh next one is what's the single most defining characteristic of Gamma's culture that sets it apart? You talked a little bit earlier around the demo culture and how fast you guys like to move and build things and get feedback, but is there anything else that really makes Gamma's culture unique? Uh, I would say that we embrace radical dog fooding. Uh, meaning that we use our own product constantly every day. This was actually a requirement for me as a founder of starting a company is I didn't work I didn't want to work on any product that I and the team could not dog food. uh which rules out like 98% of like good products you can build and and I don't think it's the only good way to build a company because many products just can't be dog fooded. They're made

for someone else. But it was a must for me. I just I I saw how easy it is to make a good product when you have that asset of your team living in it every day. So we use Gamma for all of our internal documentation. We also create reasons to make presentations because actually as a small company we don't need that many presentations. Uh for example, we have a thing every Friday called Gamarama where you could think of it as like a TED talk. One person gets to make a presentation on a topic of their choice uh to the entire company and we all then critique it and they talk about what was hard and easy about making the presentation. We are just constantly looking for ways that we can present things, write things, live things in gamma. Our Slack is just full of bug reports and then people fixing the bug reports, you know, 20 minutes later because it's bothered them too and they were empowered to fix it. So, it's

not always easy to build that culture, but if you can, gosh, it just unlocks incredible product quality. Yeah. Well, when you say that, it reminds me of Duolingo because I know famously everybody who works at Dolingo is expected to use the product every day, which you couldn't do with every product. With a tool like Gamma, where you're building presentations, you know, obviously anybody who works at a tech company is going to create and or view a presentation on a pretty regular basis.

And so, it's one that naturally lends itself to a lot of dog fooding, which is very nice when you can get it. Um, what is one thing that keeps you up at night as the co-founder of Gamma? As you said, it's an incredibly competitive space and a fastmoving space where the underlying AI model capabilities are changing completely. So, on the one hand, I worry about companies like, you know, PowerPoint and Google coming after us because they finally get their act together on building great AI. On the other hand, I worry about tiny new startups using technology we haven't even thought of yet to uh leapfrog us. On the third hand, I wonder if the AI models themselves get so good that they can just use the computer for you. So, for example, uh maybe PowerPoint never builds good AI, but OpenAI has a model that's so good that it just uses PowerPoint for you and you never have to open uh a new tool. What if we don't

even make presentations anymore because robots are doing all the jobs? It's so hard to know how the entire space evolves. And I think the best we can do is just try to stay ahead of it all. Well, and this reminds me of this debate that's been ongoing now for a few months in the tech community between on one hand people who say, "Well, ultimately the horizontal LLMs that are so wellunded and they have so much compute power and all of these resources like they're ultimately just going to end up wiping out so many of these companies in the application layer." But then you have this other increasingly loud side of the debate that says yes, but the companies in the application layer ultimately own the customer relationship and especially at the any sort of network effects either between people or between people using the product that creates this data flywheel where the more people use the product the more

usage data they get which leads to a better product experience. Those can be long-term defensible competitive advantages. And so I'm more and more bullish on that side of the argument, which is these application layer companies, if they're well built and they have a loyal following, I think they can hold off the LLM in many cases. But we'll see how it all plays out. And that brings me to my last question, which is how do you think AI will change the way people work over the next two to three years? You know, I I love this uh joke I've seen online of uh there's like a comic panel of one person using AI to write like a 10 paragraph email to someone and then the other end someone just takes 10 paragraph email and turns it down. Oh, I've seen this. uh two sentences again. And uh I do think that we're all going to we're going to find that AI just plays this sort of ridiculous glue role between all the

sorts of communication that we do. But that communication is necessary and that glue role is going to feel great because we're going to spend a lot less time doing all of the work about work and more time getting to do the actual work that matters for all of our users, customers, and everything else. Yeah, that's so well said. It will be the great leveling of the playing field where there's less and less busy work and more and more just focus on creation and innovation. Well, this has been such a pleasure, John. Thank you for coming on the podcast. Uh, it's been great getting to know you and Grant and the rest of the Gamma team over the last couple years and certainly wish you all the best of luck. Any parting words that you want to leave the audience with before we wrap up here? Working in AI is really fun and if you're not, I would highly recommend finding a way that you can get to a place where you are. Yeah.

Well, and Gamma is a great place to do that. So, I know you you all are hiring. Absolutely. Thank you, Phil.