Notion co-founder on malleable software, monetizing agentic work, reinvention, more | Akshay Kothari

Douglas Engelbart and Alan Kay talked about this. Somehow we sort of like took a couple wrong turns and ended up in this very rigid application software world we live in today, but the original computing ideas were actually meant to be quite malleable, meant to be more people to be creators uh as opposed to just consumers. In the original mission we were trying to democratize software making so that not just the developers build it. In this new world that same thing applies to AI. You just don't want only your engineers to get the power of AI. We've built these platform for people to build agents where the people team, the finance team, the marketing teams can all sort of uh build their own agents with their own context in a permission adhering way. So yeah, we've gone far away from like just being the system of record to being, as you said, like the context layer for humans and agents to collaborate with each other. And in doing so, I think we can now, instead of just selling software seats, we can now sell you work. You think about a large company, they probably will pay us some platform fee for the seat-based pricing, and they may have some usage on the custom agent as they adopt it. But you could also imagine like an AI-native startup, like five-person company, um they may pay us whatever, like 100 bucks a month as a platform, whatever seat-based price, but they could build this data scientist as a custom agent, right? And they could spend $50,000 on it. And $50,000 is still less than getting a $300,000 data scientist in SF if it can actually help them do that work. There's just a narrative around you know, AI is re- is sort of like making humans replaceable. We're just creating a lot of fear and anxiety and nervousness amongst the people, and we don't think I think AI should be about abundance, the fact that you can do more. And so like as it relates to the company, I don't think that the primary motivation for the company should be that can I do the same thing that I'm doing right now with half the people. I think it's about raising the ambition, right? Can I do like two x of what we can do with the people I have today? And can I like keep adding more people because my ambition is what I can do with the company is expanding, right? Actually, thanks so much for doing this. Super excited. Thanks for having me. Thank you. Wonderful. Um So, Notion's up. I'm going to dive in straight away. So, Notion's origin story, depending on how you tell it, is a story of extraordinary patience or extraordinary stubbornness or mix of both, right? Uh before Notion, you co-founded Pulse. Uh mobile news reader that LinkedIn acquired in 2013. And um you did spend time in LinkedIn, right, making uh the decision to start over later. Um And then multiple resets uh in this journey. And the famously, one of those things is famously locking yourself into a small two-story house in Kyoto to rebuild from scratch. So, before we get into the Notion story where it is today, I would love to hear you tell us the Pulse story first. What you built, why LinkedIn wanted it, and more importantly, what are some of the lessons that you take from that journey that you are now uh have taken forward into Notion?

Awesome. Well, it's great to be here. Um Pulse was a fun um it was actually a class project at Stanford. Uh I was my doing my master's there. And uh we took this course at the design school called Launchpad, where the only requirement was that you have to launch publicly. And so, Pulse was basically built there. It was never intended to be a company. Um and we launched it I feel like in a few weeks, maybe six, seven weeks. And uh I think a few weeks into it um in I think WWDC 2010 Steve Jobs talked about it as one of his uh favorite apps on iPad. So essentially sort of overnight became a company uh through that. Uh decided to continue working on it for a few more years. And we got to tens of millions of users um very focused on partnerships with publishers. Um Uh we had all the content, but we didn't have any identity data. We didn't know who was reading it, what their background was, and I think in 2013 we're starting to get to a point where I think the news readers were actually getting pretty good. And I think the problem to crack was more about how do you take all the content being published in the world and really recommend the right set of things that people should read. And uh LinkedIn happened to be thinking about the same problem from a different lens. They had the identity, they had the network data, and they were trying to figure out how to get into content. Uh which they had tried a few different times, but then happened to um tap into Pulse as another bet uh into the space. And so yeah, we were acquired in 2013 and um I was there for, you know, five and a half years uh really helping build uh the content platform. Uh a lot of what you see on the homepage, if you like it, you can, you know, it's the work of the Pulse team. Um And actually the year we sold Pulse to LinkedIn was the year Ivan started uh Notion. And Notion ended up being one of my first investments. Um And so that's sort of the connectivity to the two stories. Brilliant. And let's talk about the Kyoto reset. You and I we call it second act. Right? Why do you think it's a the that it was a critical juncture in the path was the one of the strongest product market fit journeys, right? That you can identify. Would love to hear that story.

Yes, so Ivan had this vision of sort of democratizing software making. This idea that why is it that only 20 25 million developers get to design software and then the rest of the billion knowledge workers just have to use that software. He sort of imagined a world where more people would be able to sort of modify or create their own software. And I think the early days, you know, we built this platform like this dream platform where you can build any software through the building blocks we offered. But we had no users. And so so that was sort of the initial sort of dream was not we were not able to like will our way into sort of product market fit. We also made a bunch of like technical architecture decisions that were incorrect. It was not very stable. And so like I don't know a few years into it we actually ran out of money. I think Ivan took a loan from his mom. And it was a really funny story where we put the San Francisco office on the Airbnb. And we were making more money on the Airbnb here than the Airbnb we were renting in Kyoto. And Ivan What year was this Kyoto reset? Which year was this Kyoto? I think this was 2015 2016. Yeah, it was basically like the last ditch attempt uh, to see if we can give it this, you know, final shot of this vision we have for the world. Uh, and I think maybe the learning was that instead of trying to build this overall platform um, we started with a Notion 1. 0 specifically about docs. Docs and knowledge base as a wedge into the market. Like we still wanted to create this platform where people can create software, but we needed a wave to ride. And the wave for us was really building a modern editor and a modern knowledge base. Beautiful. And this particular thing that you're referring to is a block architecture. Right? It's a fascinating story. And you made this as a foundational bet, right? Like almost like appropriately sync engine uh, and a block-based canvas, right? And at a time when like I think most people wouldn't have understood it. Like I don't claim to say I understand it exactly, but uh, and especially not by customers, right? And so what does it take to hold a conviction like that on a bit like this where there is no external validation to point to say this is how this pattern exists? How did you hold on to that conviction? I do think there's a little bit of

stubbornness that this is like the right model. Uh, I think what we were trying to do is essentially make everything modular. Uh, including a line of text that you write you can turn that line of text into a header or into a page or into anything you want. And so everything you use in Notion is everything is a block. And you can add blocks and you sort of create a graph of blocks on a page. And the idea was this is somewhat of a like a programming environment that is hidden underneath the product. And the hope was that the way you uh experience the product feels as familiar as Google Docs, but underneath there's actually a lot of power built in. And so, again, we started with the editor. The next thing we built was a general-purpose database, right? And the And general-purpose database meant that you could essentially use it for all sorts of use cases. Like the same thing that you use for project management is can also be used for CRM, application tracking system, can also be used for your home tasks, right? And so, we tried to build these general-purpose blocks, uh general-purpose sort of building blocks that we gave to the users. And I think the thing that actually really hit product-market fit was we took our own blocks and we created simple templates. So, we said, "Okay, you need a to-do list, we'll give you a to-do list. You need a project management Kanban board, we'll give you that. You know, you need a CRM, we'll just create a template. " And I think that sort of really put Notion on the map was people came in for all these different use cases. They started with that use case, but then when they dug a little deeper, they realized it's very extendable and they can you can kind of like make it your own. Right. That That's what you're referring to as the malleable software? Correct. Beautiful. And uh most companies, right, tend to go into a more opinionated, more vertical kind of approach, right? And what made you sure that you wanted to stay the course of a horizontal approach betting on the technology first? And how did you guys think about the PMF especially through let's say the ICP lens of, "Okay, this type of users are actually able to get it and they have a very unique set of use cases that's so different from, let's say, something like a Google Docs where you are one of the best fits for this. " Yeah.

Yeah, I think I It's not so much we're trying to create the surface level as simple as possible, but then there's a lot of power built in. I think a lot of the POV also came from like our own um values and mission about um I don't know. Like I think there's a lot of people who use Salesforce and if you want to change like one of the properties you have to call a consultant and then they have to do all this work to do that and that just felt a little bit wrong to us like it felt like most of the things people should be able to change and modify even if you're not creating the whole thing. Um I also think there was a something about empowering more people to be creators. Uh I feel like growing up there was always like two people in your class who are not only learning Excel but they all they can also do Excel macros and when people build these macros they there's a sense of pride Right. where they feel like it's not Excel it's me like I build these macros. And [clears throat] I think some of that actually like plays out in Notion, right? It's like a lot of people have built incredible workflows in Notion like incredible templates and and I think they have told so many more people about what they have built. Uh and we love that. Like in Is it bragging rights? Yeah, in some ways we're not trying to be cool we're trying to make these creators cool and sort of democratize what they can build because they were going to bring other people into the platform um to do it and it's actually you know some of a lot of these things actually not new ideas like um it it actually borrows quite heavily from ideas from the 70s like Douglas Engelbart and Alan Kay talked about this. Somehow we sort of like took a couple wrong turns and ended up in this very rigid application software world we live in today but the original computing ideas were actually meant to be quite malleable meant to be more people to be creators uh as opposed to just consumers. Beautiful. And where did you eventually end up beginning with Ivan and what Ivan has lovingly called it as sugar coated broccoli?

broccoli? Would love to hear a little bit more. I think like you know, part of it is just like I think you have this vision for the world. You sort of have to figure out like what's the wedge into the market. Um there's that great movie. I don't know if you've seen General Magic where you know, they got the best of the best people in the same room trying to build a thing that looked like iPhone in the mid 90s and like they just could not will their sort of put their you know, full vision into the world just because the world was not ready for it. The technology wasn't there. And so for us I think what I would means by this, you know, it's like we can't tell everyone to sort of create your own software. It feels a little bit like work and so we sort of have to give people like level one of the video game which is you know, maybe just take notes. That's if that's what you want to do like that's a good starting point for you. And then like we got to like get them over to level five and level 10 when they can do like more complex things with it. Um and so I think it's his way of saying that we sort of have to like pull you in before we sort of have you eat broccoli. Beautiful. So, switching topics. Um you have been very clear about what's actually shifting, right? More than most founders which is you named the real threat early which is not white coding not the model makers eating your lunch directly but the question of whether your software can exist inside an agentic experience at all, right? So, the next battle of software enterprise software is increasingly described as battle of context and by design, right? You hold so much unstructured data than most tools, right? And that's a phenomenal advantage. But the context that an agent actually needs should also work across multiple systems. So, how do you guys think about your vision in this type of a world and uh would love to understand how you think

about the connective tissue around all of this. Yeah, we I think we definitely consider ourselves quite lucky that we bet on AI pretty early on. I think we released our alpha 2 weeks before ChatGPT came out and I think it was sure luck that we got early access and we leaned into it. I think what we see today is actually quite interesting, you know, we talked a little bit about sort of what we've built is like the software platform where people are using the software to organize their work and run their projects and so forth. I think what AI's now enabled us to do is not just sell software, but actually sell work. Like people are able to just essentially say like here's the types of things I need to do and it can go sort of execute on that work. And increasingly I think we're all going to live in a world at least for the next decade where we're going to work alongside agents. And um this is already happening in the world of programming, right? I think like my co-founder Simon, who's I think one of the most prolific coders I know, hasn't written a line of code in the last 6-9 months, right? Um in instead he has five agents running in parallel doing work for him. And I actually think that is going to happen with other knowledge work as well. Like I think all of the our roles is going to move from us doing the thing like the way we do it today to more of a factory where we can run some of these agents. And so if you think about that world, agents can only be so helpful if they have the context. Like we can do certain amount of work because we have all this context that we gather. And so increasingly Notion has basically become the infrastructure that provides this context in a permissioned ideal way to these agents, right? And we just are sort of lucky that we built a system of record too, right? So if you write a doc and your team's writing a knowledge base and all your information is in these databases, all of that is useful context that the agent has to go do the work. Right. And the way the agent does the work, it can also create documents and create workflows that you can then inspect, evaluate, approve, so that they can continue doing that work. Um and so So, yeah, we've gone from away from like just being the system of record to being, as you said, like the context layer Right. for humans and agents to collaborate with each other. And in doing so, I think we can now, instead of just selling software seats, we can now sell you work, right? We can run automations for, you know, like we can make sure like all your AEs are ready with all the research for all the customers they're going to talk to. We can make sure all your product feedback is routed to the right teams and bugs are filed and bugs are sort of worked on through coding agents and so forth. Uh so, it's actually quite an exciting time to just think about what the power of coding agents has done to the world of soft- software programming and then sort of make that available to the rest of knowledge work. Brilliant. And I want to pin the topic on the user, individual user and the association with pricing. But before we go there, I do have one more question about the model makers, right? The model makers typically are forward integrating, right? In the case of Microsoft and um Office and with Google and vertical search and now it's happening with AI labs. And when Anthropic now ships uh productivity plugin, every horizontal knowledge tool has to defend itself, right? In the case of a Thomson Reuters, they deployed um the their answer is 2,700 attorneys attorney editors and decades of curated legal content becomes their answer, right? Yeah. How do you think about your approach um and what advice do you have for fellow founders and builders?

Uh especially when they are building on top of these models. Yes, so there's a couple of things that I think that's helped us especially as we think about the world of enterprises, right? Like what do they want? Um a lot of the AI tools that exist today are very single player mode. Uh If someone's trying to like transform their company, like if you just give everybody chat GPT or Claude, I think the best people will find ways to be very productive, but it's sort of lose you lose a lot of people who don't know what to really do with it, right? And I think what Notion provides is this multiplayer platform where a single person can actually build an agent that can be widely deployed inside the company in terms of really driving that productivity across the company. Um building an agent is as easy as collaborating with you on a doc, right? We can co-build it, uh control the permissions we give to it, uh write a job description of what this agent does, and then let it run on Notion today. And so it's collaborative in how we build it, but also the broader company gets to experience it, right? In terms of it answering questions in Slack or it running some actions on other platforms and so forth. So that's one piece. The second piece is like I think the more we talk to companies, the more companies do not want to sort of pin themselves down to a singular model or singular company. Uh and I think this is where Notion has increasingly become the Switzerland of models. Like you buy Notion and then you get access to all the models including the big labs, but also open weight models. Uh and then I think the last one is most importantly, it's sort of like back to the mission, right? Which is in the original mission we were trying to democratize software making so that not just the developers build it. In this new world, that same thing applies to AI. You just don't want only your engineers to get the power of AI. We've built these platform for people to build agents where the people team, the finance team, the marketing teams can all sort of uh build their own agents with their own context in a permission-adhering way. Beautiful. And so that's sort of what's distinguishing us. As it relates to other companies, I I still think it's so early. I think a lot of people sort of look at the crazy scale Anthropic has and it's like I think I don't know if we're going to live in a world where one company will rule everything. I think there is um a lot of like hard infrastructure to be built. Right? I think if you just think about um like we think a lot about just like okay, well what is the cloud and what is the database in today's world, right? All of these things were built for humans to operate, but now you're going to have like millions of agents. And so you have to completely rethink what this infrastructure will be. And just a lot of opportunities. It's still so early in this whole journey. Very true. Thank you. Um switching [clears throat] topics. Now, uh on pricing. Yeah. Um can we — My favorite. Our store. — And so when we featured uh Brian uh from Notion who heads growth product, um this is for our annual research report. He noted something um particular that stuck with us, which is how Notion's pricing goal um as with everything else has been centered on simplicity. Mhm. Right? And it is always reflected in your C-based pricing, right? And the world was much more simpler then. Right? And uh Notion has been constantly adding AI features uh over time and including custom agents, right? And then something seems to have shifted where the custom agents seems to have done something to the way you think about pricing. Now it has introduced the usage patterns that were like the variables have been introduced. So from you held a simple fixed model through the years of AI development, right? And what was it about custom agents specifically that made the old policy untenable, right? And uh if you can elaborate on that especially for the uh through the lens of others who will be building agentic products and introducing to the market. Your learnings will be very valuable. Yeah.

Oh, we can spend another full hour on this. Uh I'll try to be brief. Uh we tried very hard to keep it simple. You know, I think part of it is because I think um you can tell a lot about the company through their pricing schedule, right? It's like um So for us like, you know, we had the typical good, better, best seat model. And when we first introduced AI pricing 3 years ago, um in the early days of AI, everybody was charging on tokens. And when we did that research with a lot of people, people realized they were just very stressed about every time they would use the call, right? And so we came up with an early version of the pricing which was a just an uh subscription model. You pay $10 a month and you basically get all the AI stuff uh unlimited. Kind of like, you know, based off of the cell phone plans we have. Right? There's some overage policy, but like most people don't hit it. Um so we did that for a long time and then we realized that was a good starting position cuz some people were choosing to be buying the AI stuff and some people didn't want to do it. So that made sense then. Then roughly about a year ago we realized like most people want this AI stuff. Like why do we have it as an add-on? So we removed the add-on and we put it in our better and best model, right? So you can still buy the good one that sort of gives you very basic things and then better, best gives you full AI. Um I personally prefer like the seat model just easier for CFOs to like understand, like it's predictable. Like you know, all of that stuff is true. The part about agents, specifically custom agents, is that the um that the range of things you can do with it is quite broad, right? You can actually use a very, I don't know, simple model, cheap model to triage your email, and write draft responses, right? Every time it runs, it's probably like, I don't know, some pennies to do that. Uh but you can also uh create a custom agent that can write queries on top of your Snowflake tables, and um it's like as good as a senior data scientist. And each query could be like $5, right? Cuz it basically uh does a lot of uses a lot of coding agents to do that. Um And so we came up with all sorts of different ideas for like, okay, how do we price this thing, right? And the challenge with the fixed fee or like a seat model is that you will always be forced to limit how much people can use with it, right? You'll always be like you don't want to be have negative margins, or you just whatever the existing margins are, you're always going to be hitting up against that. And it just felt something like um where we are with what people can do with agents, you sort of have to create uh figure out, okay, how do you price work, right? Uh how do you give people intelligence as a service, right? And even though it is charging per activity, it's closely tied to the work you're doing, right? So like if you have a mail triaging agent, which is just triaging your email, which I use every day, um it probably can cost you like, I don't know, 10, 20 dollars a month. Uh and then it's up to you like is that worth that much, right, to you? And then you can also choose between the models again. Open weight models could be lower. You could also use Opus and maybe spend $100 if you really care about that quality and so forth. Um and so we're essentially giving people this we decided to separate it out as a separate product. Custom agents, you set it up, set the model, and then you're just paying us for the work you're doing with it. Right. And we released it last month. Uh we made it free for 2 months so that we can adjust all the feedback we got. Uh but we feel pretty good about the model now and so I think um we're going to enforce it in about 3 weeks. Um and I think I you know, we're keeping the seat model as the platform, right? So, if you're using the AI writer or some of the older AI features like enterprise search, all of that is available in the platform fee. And then if you're using custom agents as a way to um do work in the future as a way to potentially have uh a custom agent be doing all the work like a full-time employee is. — Right. Something more sophisticated. — more sophisticated. You can decide if you want to sort of pay for that and we don't want to limit you with a seat-based fee. So, which means that you're still going for adoption and adoption of democratizing AI capabilities and AI agent capabilities while and that is part of the platform fees and then you're actually layering this with like a sophisticated layer that would be usage-based. Exactly. So, if you think about a large company, they probably will pay us some platform fee for the seat-based pricing and they may have some usage in the custom agent as they adopt it, but you can also imagine like an AI-native startup like five-person company. Um they may pay us whatever like 100 bucks a month as a platform, whatever seat-based price, but they could build this data scientist as a custom agent, right? And they could spend $50,000 on it. And $50,000 is still less than getting a $300,000 data scientist in SF. If it can actually help them do that work. And so for for the CEO, it's like an it's a decision whether they can connect all the data tables, connect the right permissions, and enable an agent to go do that work. — Beautiful. So, this is uh a different way of looking at hyper-localization, where there is a purchasing parity that actually works differently when you think about custom agent, depending on who is your customer globally, right? And they may assign different value, right? To that. And the ones that choose it differently might actually be opting in. Yeah. It's a beautiful And I think this is where also the companies having the controls of which models work on what agents, and that flexibility of being able to give the right context, everything matters a lot, right? Um I think if you choose a specific company, a specific lab to do all of your work, I think you may end up with something that's 10 times more pricey, right? Whereas if you have that choice to go try all the different models, including the open weight models, um I think you can significantly do much more work with a much lower cost. Beautiful. I like it because uh this is uh reversing the Google's famous hyper-local pricing, Yeah. right? Where they do it based on purchasing parity, right? Is very different. But here you're actually flipping it where it's through the lens of value assigned for unit of work by the customer to think about like opting in for a custom price till I think the market matures for us to be able to figure out like where it uh settles. You know what other thing we realized in just the whatever like 40 days we've used like is people like these agents have gotten so good at writing like simple code, right? And so in some ways you can actually use the GPUs to write code that can let's say sync a Notion database with like Salesforce database, right? You can have GPUs write code to sync that. And then once the code is written, your cost moves from GPU to CPU. Right? Because you don't need a GPU to like sync it every time. Right? And so, there's all these interesting things we'll be able to provide our customers in terms of cost advantages um as opposed to like really throwing an LLM for everything. Very interesting. Now I see why you said we can keep talking for an hour.

— Okay. Uh switching topics. Um what are some of the North Star uh metrics that you are tracking now? Especially, how have your internal growth frameworks changed as a result of thinking about like what you track and what how has it changed? Especially the North Star metrics. I think a lot of it remains the same. I think it's, you know, sort of always um sort of leading indicator is always usage. I think um maybe in the previous era we used to think about monthly active users or daily active users. I think the new dimension you have to think about is like agentic usage now. Um right? It's like do you have real activity, real usage flowing through your ecosystem? Um and then of course you track um you know, revenue and ARR and especially right now it's like net new revenue we're adding, you know, every week, every month is something we like track quite religiously. Uh and then the thing that has actually kind of evolving a lot right now, which I'm sure you'll probably think about is just margins, right? I think there's a world we lived in where just 90% margins was like the norm in SaaS. I think a lot of that stuff has just been thrown out the window, right? Uh and it's almost like if you were a 90% margin business, people perceive you as not having any AI usage, right? Uh and so I think it's kind of an interesting world where people, especially the street, really cares about you being a platform and seeing real AI usage happening. Uh and in this, you know, in this sort of um usage-based pricing world of custom agents, like, you know, we're we're obviously capturing some fee for processing and setup costs and so forth, but we're nowhere close to the SaaS margins that we're operating on. Yeah. Beautiful. So, uh switching topics, the the particular narrative, right, the prevailing narrative about AI and the future of work, um runs in a that is work runs in a particular direction, which is fewer people needed, leaner teams, and eventually even a billion-dollar one-person billion-dollar company, right? The headlines have a certain flavor to it. Um and Notion's answer to it is different. The bet you're making is definitely people first, right? And uh as the session's titled, right, as we put it, that is that AI's highest use is not replacing human collaboration, but expanding it, right? And that the teams and organizations who learn to work with AI together will run circles around those who treat AI as a substitute for people, right? And uh that software should remain at its core a tool in service of human thought, and uh shaped around the brains and not the other way around. And um where does this particular like case come from? And I'm I want to know more about like the kind of discussions that you're having internally, right, about this. And more importantly, what are you seeing power teams do on Notion alongside agents that is reflective of how this collaboration could look like where the how the collaboration at scale wins over like the radical individual leverage idea, right? Would love to hear a little bit more on that. Yeah, so it's a super fun one. I mean, it's definitely comes from our heart, right? So, I think there's um I think on one side there's just like these headlines about like when are we going to have a single person running a

billion-dollar company and I don't know, like for us like that sounds so lonely. Like I don't want to live in a world where I'm just working alone. Uh I think some of the best work we've done uh has been groups of people coming together and solving that problem. Like that is the fun part. That's why I go to work. Um and so um so that's one. I think the other one is like also like I feel like there's just a narrative around you know, AI is sort of like making humans replaceable, which is creating a lot of fear and anxiety and nervousness amongst the people and we don't think I think AI should be about abundance, the fact that you can do more. And so like as it relates to the company, I don't think that the primary motivation for the company should be that can I do the same thing that I'm doing right now with half the people. I think it's about raising the ambition, right? Can I do like 2x of what we can do with the people I have today or can I like keep adding more people because my ambition is what I can do for the company is expanding, right? And so um so it very much sort of comes from that point of view of the world. Um it's an interesting sort of if you if you go back in history, I think there's like two schools of thought. There's sort of this like AGI school of thought, which is sort of you can replace humans and like, you know, machines are as good as humans and so forth. Um There's another school of thought that sort of Engelbart and Alan Kay came up with, which was uh not AGI, but AHI, which was augmenting human intelligence. It's sort of just like this concept of bicycle for the mind, the fact that we can all be doing much more. So it's like kind of like humans plus AI can be doing I don't know, so much more now, and uh we think that's the future, right? It's like it's not I don't think even like the world cares about efficiency is like the thing, right? It's Whatever, yes, you can get some credit for it today. But the fact that you can grow your ambition as you grow five times faster over the years to come is what is being rewarded, right? It's like an average SaaS company today, I just checked, the mean and median is 11%. Growing to like 12%, right? And we're 3 years into AI. Like 3 years since ChatGPT was born, right? So it's like why is it that the median company is can go from like 10% to 11%? Like why is it not going from 10% to 30%? Right? I think that's the question all of us have to ask. Is not that we can be like whatever, you know, 30% operating margin versus 10%. It's more like why is your growth rate not 50%? Why is it not accelerating in this world where anything can be built? Beautiful. Beautiful thoughts. And that brings us to the narratives, right? Which is uh now you are personally leading Notion's storytelling team, right? And we are borrowing the idea, by the way. Uh and you have run dozens of public demos, right? And including I loved you and Ivan coming and doing the demo for us, right? For me and Raman, and we just loved it. Again, another idea we are copying. Uh shamelessly. — we we've borrowed it, too. — And you have launched a Notion media fellows program, right? And um love the uh think together clip, right? So, would like to tell us how the tell us a story of how this face of your role and the current focus on telling deeper stories at Notion has come about and why now? And if you can tell a little bit more

about that, it'd be great. Well, yeah, some of it what we talked about. I feel like it in some ways it felt like the narrative of the world today was a bit gloomy to us. Like I think we just wanted to inject a bunch of optimism. That's why we did the think together campaign. Um is because I don't know if it's like an interesting history. I think in the 30s, IBM came up with a think campaign. The think campaign back in the 30s, the enemy was like whatever uh people feeling afraid of computers or whatever, right? And then in the 80s, Apple came up with think different. Uh and that was really sort of thinking creatively out of the box, kind of like you know, the misfits, right? And think together for us was a little bit of a ode to that, which was um groups of people coming together and building things together, right? So I think the response of that was really incredible and interestingly like from idea to execution was about 3 4 weeks. Like we thought of this and we sort of built the video in-house and we pushed it out. So that's one piece. I think the other piece just on the business front, like I just feel like I don't know if you feel this way, but as a product person I just feel like oh our product is so far ahead of our story. Right? And it's part of it is because I think the company has been around for a while. And so people think about Notion, they think about oh yeah, that note-taking tool. But it's like no no, there's so much more. Like all this platform we've built, all this power the agents we're providing and so for me personally, I just felt like instead of me just delegating to the marketing team or the comms team and complaining that the story is not being told. I said, "You know what? I'm just going to take the microphone and start doing demos publicly myself. " And I feel like it sort of like broke the the rigidity that existed inside the company where I feel like as we grew to a thousand person company, everybody was just looking for permission. It's like, "Oh, can I talk about this? that? Am I leaking this? " And then when you have a founder just leaking stuff all day long, then it's just like, "Okay, well. " And I love it. I think everybody can do it, right? And so and I think it's like I feel like more people got to know about the product, what it can do, how it can transform the company through these like really scrappy videos that I was shooting at home and office. Um And it to me was just a little bit of like, "Okay, I got to lead from the front, show how to do. " And now like there's I don't know, 50 people inside the company, engineers, designers, marketers, sales people who are all doing demos and I haven't done it in a while because I'm like, "Okay, the the train is, you know, moving. " And it's actually kind of a fun It's been a super fun 6 months thinking through it and we're far from done. I think we got to do it more. And actually some of this like levity, just lightness existed when we started the company. Right? It's just you add a bunch of people and just you get a bit siloed, a little bit rigid and I just wanted to break that. Right? I just wanted it to be like, you know, it's like ask for forgiveness, not for permission. Brilliant. Love that. So, one last question. And something that I always ask is there a particular book that you have gone back to again and again? Um especially have as you have navigated this particular exciting stretch of the founding journey and also now it's almost like reinvention of the company. Like are there any particular

books that you recommend? I'm a big fan of just Warren Buffett and Charlie Munger. So, you know, I've been going to their conference since 2014. Um It's like a you know, especially when you live in the valley and everybody's like 5x, 3x, 5x. Like just kind of like grow, grow. I feel like it's once a year it's like going to the temple of long-term compounding, you know? — Where these guys are just 20% compounding for 60, 70 years. And it's always like a good reset for me. And so, I go back to their letters. And I think, you know, also the Almanac of Charlie Munger. Because uh it's just like you know, they a lot of times they were just talking about sort of giving you this um wisdom on financial stuff, but it's like a lot of life advice. Like so much of it was just life. Like just living a good life. Poor Charlie's And Yeah, and to me it um yeah, like I just love the values, the way they live their life, the way they sort of educated the world. Um and you know, you could I don't know. Just pick up a random letter from the 70s or 80s. Just how Buffett was thinking about the business then and there's always something interesting that you can find. Um so, so I really enjoy sort of going back to that. Thank you. Thank you so much. This was so much fun. Really appreciate —