Business used to be go karting. Now it's Formula 1 | Akshay Kothari (Co-founder of Notion)

The terminal value of these old businesses, you should just assume they may go to zero. You have to disrupt yourself to where the business is going. Notion is like an art project where we care a lot about the piece of this malleable software. Because of the investment in the AI, we moved from go-karting to Formula 1, and I think we have a shot at the podium, but I think we have to rewire the race we're in now. There is a real fight for survival. The old guard is falling, creating this incredible opportunity for new players to become the new guard.

3 years from now, most of the pre-AI software companies you can name today will be gone. Acquired, irrelevant, or quietly winding down. The AI-native startups will have eaten them. But there will be exceptions. A small handful of pre-AI companies will not just survive, they'll come out of this stronger than they went in. Notion is going to be one of them. By all accounts, Notion has always been a unique company. They took 4 years to ship their first product, they have more cash on hand than they've ever raised, and the way they talk about product, people, and operating is truly different. But what might make this story even more special is how they've transformed themselves to adapt to the AI era. My guest today is Akshay Kothari, co-founder of Notion. This episode is a field study in what it actually takes to make the transition from a pre-AI company to an AI-native company. Let's get into it. Akshay, welcome to the show.

Thank you for having me.

So I want to focus most of this conversation on the reinvention that Notion is going through in the AI era. But before we get there, let's talk about Notion the company. It's a pretty special company. Take me back to the founding days. What were some of the core founding ethos and principles you guys set right at the beginning that makes Notion the company it is today?

Well, one of the ideas that's really core to us is this concept of malleable software. This idea that everyone in this world can create, modify, edit the software in the way your brain works rather than having to sort of accommodate how, you know, today's rigid software works today. And it's an idea that's existed for a while. I feel like if you go back to how the pioneers really thought about it, computers and what it can enable back in the '60s and '70s, this idea existed, but somehow like it never took off. And when Ivan started the company, I think that was actually at the core, was that give people more of that power to be able to do it. So that's one of the things, which is, and I think what comes out of that is this idea of like systems design. I think that's always been very core to how we've built the company. is if you think about like, who are we trying to hire on the product side? We really value a lot of people who bring in systems design, who can kind of think of the overall system and recognize that making changes on one subsystem can affect the broader system. And you can see that in the product if you use it, it's very modular in the way it's built. The second one is the word that gets used a lot today, but I think Notion's always been known for its taste and craft. It's a product that makes you feel good, that sort of gets to the feeling. People who use the product for hours feel like they created something. It's very much like a creation tool that people are really proud of. So that's something we've injected by building a platform that people like to build with. And then I think maybe the last one is Notion's a very unique company in the way we went about fundraising. And we really sort of operated from first principles. Many years ago, I think in Silicon Valley, I think you could just raise, I think people tend to raise when they can. We sort of resisted that for a long time and in some ways we became profitable. And so what that meant for us is that the founders and the employees You know, still own vast majority of the company. In none of the financing rounds did we have to sort of put a board member. We were able to raise on our terms. And in the early days, I think a lot of investors used to look at that and be like, well, good luck building a lifestyle business. But we're a business that was, you know, raised very little money, but has been able to scale to a public scale business today. And I think Part of it was always really thinking about like, well, is the money going to change anything in the way we operate today? And we did end up raising a lot over the COVID years and other times because there was a need for it. But otherwise we sort of always thought about it, not raise because we can, raise because there's a real reason for doing it.

The craftsmanship especially came through in the very, very early days itself. The black and white aesthetic, the minimalistic aesthetic, even on your website, the founder pictures were all caricatures, right? than the pictures and probably is still true today. So it definitely showed through even to the users right at the beginning.

Yeah. In fact, I think among the first 5 or 6 employees was an artist designer, Roman, who's still with the company. He literally hand-sketched a lot of the avatars for employees. In fact, that's one of the perks of joining Notion in the early days was that you would get this hand-sketched avatar from Roman. And I would say also there's a, I think Notion tries very hard to connect the product and the brand tightly with each other. Typically, like, you know, for a lot of companies, you sort of see the product and brand both doing interesting things, but maybe diverging a bit. You don't see the connection. And I think we've tried hard to sort of keep them tight. In fact, down to the office, I think a lot of people who walk the Notion office feel like it's an extension of the product. A lot of credit goes to Ivan. I think he really cares about these things. And he's also, I think that's his zone of genius, like product design, brand design, the connection between the two. It's a lot of hours spent and it shows in the way the brand is perceived.

Were there any cultural principles that you guys set in terms of people aspect of the culture right at the beginning that have scaled well?

I think the biggest one was just keeping the team small, keeping it efficient, not because we were trying to be like more profitable, but just because we felt like keeping it small meant that each person would enjoy their job more. It will allow us to be a lot more nimble around changes. It's easier to move like a small boat than a big ship. In fact, in the early days, I think Ivan and I both dreamed that we would always keep the company to 50 people and Of course, we've sort of crossed that number a long time ago, but some of those ethos still are true even at 1,000 people today. We fight pretty hard to think about every new hire, like why we have to do it. And it goes back to systems design, right? A lot of times you add people because it's the easier path is to just throw humans at the problem. I think the harder path is to actually build good systems internally that can scale with you.

Anything in those early company design principles or cultural principles that did not scale?

I think the thing that breaks at various times inside the company are definitely like the communication methods, right? So in the early days, actually up until 50 people, we operated with a singular Slack channel. We sort of banned people to create any new channels. And so everybody only had a single channel called General. And we were trying to recreate this, like, you know, the water cooler you have in the office is a singular place where everybody's working. And we wanted to try to recreate that in the digital medium as well. And so you couldn't create a channel and everybody was on general and everybody's on the same page. Of course, we couldn't scale that. And so now we have millions of channels like everybody else, but in those early days, it was quite interesting and effective in keeping— things in one place and like everything spawned off Threads rather than having different channels.

In the early days, the team is everything. How did you guys go about hiring your first 10, 20 employees? Was there a lot of intentionality there also?

Very much so. And that intentionality sort of deliberation still exists today. I think up until 400 people, Ivan and I met all of those folks, either both of us or at least one of us would meet as the final round. One of the superpowers founders have is the ability to say, know, like ability to veto. And I think Ivan would leverage that often. And it was a way of him nudging all of us to look for great and not settle for good people. And I think that's one of the things I've learned over the years, especially in the early years compared to my previous startup. You as a founder have the superpower of the talent density you want to create is very much in your hands. Ivan was willing to take the pain of Like, even though this, we need this person yesterday, he was willing to not settle for good and maybe wait out another month until we find someone great. And that meant in the early days, keeping the team quite small. I think we operated at scale with like, you know, a handful of engineers and 2 designers building the product quite a bit. And even today at 1,000 people, we have under 20 product people, under 20 designers. that sort of design and sort of run a lot of different product surface areas.

Were the first 20 people from your extended networks? Were you, how were you kind of finding these people?

That's a great question. First 20 people were very much part of this subculture called Tools for Thought, or people who were interested in computing history, people who were fans of Alan Kay and Douglas Engelbart. people who are believers of malleable software, I think they almost sought out Notion because very few companies are working towards that vision that folks like Alan Kay had built. And so Notion was maybe one of the early companies that actually was almost chasing this research idea that also was a decent business.

Was this subculture existing on the internet? There were groups that would meet around this or online at least. How did it actually play out? This is super interesting.

Yeah, I think there's a lot of people who are really interested in the space of tools for thought. There's some amazing writers, there's amazing engineers who have been following this space. And for a lot of them, we'd see them at events. A lot of them, we sort of follow them on Twitter. For example, Ivan found Simon, who's my third co-founder, on Twitter, just following the work he was doing. And Simon was in college, so he dropped out of college to join Notion as an intern. And then he never went back. You know, he went from intern to being Ivan's partner as we rebuilt Notion many times over. And their work and their Twitter presence was followed by other people who were interested in this space. Even today, I think, you know, there's people in our company who joined in the last year, folks like Max Schenning, Jeffrey Litt, and Mary Cook. I mean, these are folks who, you know, we used to respect a lot over the years because they were in the space. We admired their work, whether it's like building a company or research, and we have the delight of working with them today now inside the company. And so I think Notion sort of, if you're interested in that space, I think it's maybe the best company to work at.

That's amazing. So you talked about how you and Ivan met everyone till 400 employees, but today you're 1,200. So how does it work today? 400, it seems like it started breaking that both of you can't meet them.

Yeah. So I think just physically meeting or meeting on Zoom became harder because we were just hiring way more people. And so if you had to do that, I think you'd basically be just meeting them every week. Like most of the week would be spent with that. So what we did was we created a process where there were folks who were doing the final interviews. Some of them were part of the leadership team. Some of them were managers we would trust, started to do the values interview for us. And we trained them, like, you know, in many interviews, they were just joining us to shadow us to learn that. So that was one of the things we did. And the other thing we still have, even at 1,200, is that after all of the interview process is done, the hiring manager and the recruiter has to create a packet that Ivan reviews even today. And in that packet, you're going to see the scores, the reviews, the reference checks. And for a lot of them, Ivan will, if he knows people, he will do a quick background reference check just to make sure there's some checks and balances in there. But I feel like even if everyone wasn't doing it, just the fact that if you're an IC or a manager and you have to follow that process introduces a level of rigor where people think twice about recommending someone needs to be hired at Notion. And so even though we're not doing the final interview ourselves, I think that process is really working for us.

It's interesting you say that because that you've kind of tried to have now a group who can do the values interview because I was talking to Ashwin at Decagon and they had the same thing that him and Jesse would interview everyone, but as the company has scaled, they've created a group called Bar Raisers for the interview process. And it's a select group of people who make sure that everybody joining will raise the bar at the company. You have something similar, it seems like.

Yeah.

Yeah.

The Bar Raiser program that I think Amazon had for many years, I think it's fantastic, right? Because it's almost like there are these people who are not even part of the function you're hiring for, but they can particularly look for specific types of things you're looking for, whether it's values, whether it's talent. And you need that neutral sort of score to sort of think about like, are these people going to be the right fit for your company? But I would recommend founders try to meet as— we stopped at 400, but I think if you can delay that even more, even better.

Yeah.

instills a level of discipline. Every hire matters. Like by the time the person's walked into the company, they've already spent some time with you, with the founders. And that has a different equation. I wish I knew all the 1,200 people who work at Notion now, but it gets harder.

You've said before that you prefer having zero reports. How does that square with you being the COO of a 1,200-people company? That preference versus where you are in your role?

Well, so I joined, you know, very early and my job was essentially to do sort of everything except product. Ivan was very focused on that. And so I was hired to essentially like scale the go-to-market, scale the foundational orgs and scale the business in the company. I had spent many years at LinkedIn. You know, I think by the time I left LinkedIn, I was responsible for the international R&D org that was maybe just shy of 2,000 people. And so In some ways I had sort of experienced that and had no desire to like, you know, build a bigger empire and had no desire. I had no pride in like having a bigger team. And so in some ways I sort of prefer almost to have the smallest team possible. Right. And so the way that it sort of plays out in my view is, I guess like I've done this many times over inside the company is that I sort of run after the problems, sort of the biggest problems we have inside the company. I try to do that job myself first. Once I feel like I understand the job, then I spend time hiring the next couple ICs. And actually through the recruiting process, I learn even more about that problem or that function. And once I feel like, okay, we're doing the right thing, we've built the right system, I have a product mind. So I feel like every function I sort of thought, think about, okay, have we built the right system? And now is the time to scale. At which point I try to hire a leader and nuke myself out of that job. And I've done this many times over. There was a year, 2023, where I had zero reports. And I think the reason was because initially I was hiring go-to-market people, then I nuked myself out of that with a CRO, and then I did the same for people, finance, and legal. And after I hired those people, I thought the best way to hire those people or the best way to get the best people would be people who report directly to the CEO.

Mm-hmm.

Like they don't have to report to me. And I was very happy with that setup. And so I think by the end of 4 or 5 years at Notion, all of these leaders were reporting to Ivan and I had no reports. And so I went back to being an IC and it was actually a good year to be an IC. 2023 was like the early years of AI.

What was your role as an IC at the time?

I was essentially like, so Simon is also an IC. He was essentially building the technical side of sort of how do we think about AI? And for me, I was just trying to think about like, okay, well, how does the business evolve? I was just meeting a lot of different companies. I was spending a lot of time thinking about M&A in terms of talent we wanted to get. I was thinking about how we price these things. I was talking to a lot of labs and really helped shape sort of the early versions of the AI product. It's actually kind of crazy underrated to have a founder that is an individual contributor. Because essentially for me, like I am, my only optimization function is Notion winning.

Yeah.

Like I don't care about who reports to me. I don't care about any of the underlying sort of bureaucracy or politics companies have. For me, like I'll just run after the problem I see companies facing and I will try to fix it and then I will hand it over to the right person who wants to lead it. Actually that year was really productive for me, but also for the company because We made a lot of progress in AI. I had no one-on-ones, no performance reviews, no, you know, like literally my calendar is free to build and to think about how we will evolve as a business. And we even did a couple of acquisitions during that time. And so it really helped me think about like where the overall sort of company is headed. I came back to a job. I actually sort of did a full circle and then ran product for the last 2 and a half years. We'd spent a year trying to find a CPO. And then we thought it would be hard to get someone external who would be able to do it. And so Ivan asked me to take that role. So I did that for about 2 years. And then I think just in the last month or 2, I've given those Legos to Max as well. And so like I'm slowly getting back to time, which is why I'm able to do this too.

What does your collaboration with Ivan and Simon look like? If it has changed with AI, because of the initiatives, maybe pre-2023 and then post-2023?

Yeah, so Ivan, Simon, and I are all product people at heart. And so in some ways I feel like we all 3 of us sort of resonate, but I think we just have the perfect complementarity in terms of like, one analogy that comes to mind is I feel like in some ways Simon is the architect, Ivan is the sort of chief interior designer, and I'm the contractor. I think of it as like Simon's very much sort of the technical brain. Ivan's very much the sole design vision of the company, and then I am the intersection of product and business. And so I think the 3 of us actually work really well together because we have those distinct roles. Ivan is the CEO, so if we disagree, I feel like it's very easy to just follow what Ivan is saying. But I think all 3 of us care a lot about the broader company winning. And so, you know, a lot of our talks are of course about product, but we also spend a lot of time thinking about different functions and how we need to evolve and what we need to do and areas we're lagging and what we need to do to sort of improve that.

Have there been occasions where the 3 of you feel very differently about the future direction of Notion?

Interestingly, the direction we're all pretty aligned on. Where we differ is back to the time window that we're operating on. I feel like in some ways Simon's operating at 1 to 3 years out timeframe. So he's definitely the most AI-pilled. person. He can see where it's headed and he sort of works on the frontier edge. Ivan's a little bit more midterm, like he can see that, understands that, but he also sort of thinks about like, okay, what do we need to do for the next midterm? And I'm sort of more focused on today.

Yeah.

So I think, I guess if we ever disagree on something, it's less about fundamentally we have a different view of where Notion is. It might just be a debate of like, okay, the view that Simon has, is that applicable today or do we have some time to like evolve the company towards that future?

Notion does feel like a very intentional company in every aspect, how you hire, how you think about product and the company itself. Does that intentionality ever come in conflict with speed and urgency?

I think maybe a couple of years ago the answer could have been yes, but I think as of today, like if we've gone back to thinking about Sheer leverage small teams can drive, right? And so maybe in the past it would have been like, oh, the product's working, you need to hire like, you know, 300 salespeople and that's the only way to sort of scale the business. I think today, I think people are asking if that's actually the right way to do it or whether you want to like build the right systems, connect your go-to-market tools better, get sheer leverage so that like, you know, each sales seller could be like 5 times more productive.

Yeah.

I really feel like the deliberation can help you go faster in the long run, but it may actually slow you down in the short run. And as long as you have a very time, very long time window for what you're operating on, I think building systems is sort of the right thing to do.

It also though comes with the way you've set up the company to have long horizons, cash flow positive, all the things you set right at the foundation. Many other companies operate on a treadmill of fundraising. And so then they have to meet short-term milestones of growth.

Yeah, it's been a huge unlock for us in retrospect. I mean, during the time we were doing it, I don't think that was like really on our minds. If I look back at like how little time we've had to spend on fundraising or investor updates or, you know, a lot of these other things, each minute of that time has been reinvested in building a better product, selling to more customers, growing the business, you know, continuing to stay cash flow positive. I think there's a little bit of just like weight off your shoulders. Like you just don't feel like there's some time window when the money runs out. We've just never had to deal with that. And it also means that I feel like in some ways all 3 of us are artists at heart. And so we can sort of build the company in a slightly more romantic sort of future-oriented way.

The way you describe it, it feels like a craftsman at work kind of a company. Does it come at a cost of sheer operational intensity?

Again, like if you take a very long time window, I don't think so. If you want to build a company that is going to be around for decades, I think you want to keep that intentionality because those values, those culture is the thing you'll scale. I think we're in this funny period which we've experienced before where it's just like people are just trying to like 10x, 10x, 10x. And it's incredible that some companies can do it and do it sustainably. But I think there's a question of just like the durability of that. How are you sort of like thinking about the culture of the company? And then like, what exactly are you trying to scale? Right. It's like, are you just trying to do this as a way to like maximize the valuation or sort of the financial gain? Or are you trying to like create the thing that you really, really deeply believe in? Right. It's like Notion is a very much like, Yeah, as I said, like an art project where we care a lot about the piece of this malleable software, which by the way benefits a lot from AI, right? So in some ways I feel like you had to put these pieces together to create something. Now you can ask the AI to do it. The great news is like what AI creates is not something rigid. Like it'll create something that you can actually modify. You can feel, you can kind of like move things around. you don't have to keep prompting your way to it. And so there's this interesting balance of it's easier to get started, but it still has this benefit of being able to modify by hand.

Yeah.

And I think that's our unique point of view of the world. And I think we are trying to scale the usage and impact of that in the universe. And once a year I go to the Berkshire Hathaway conference and I sort of feel like it's like once a year going to the temple because you are around people who are just thinking in decades, thinking of compounding and what that can do. And so for me, it's like, yeah, like I feel like maybe there's always this urgency of like, oh, should we be growing faster every year? But it's also like, okay, can we just do that? But for decades to come.

Are there aspects of Notion's culture that you sometimes feel like Other founders should just steal almost, right? Like it does feel like a special company, but you're also an N of 1 company. Why are more people not doing how you guys think about things? Like what are the 1 to 2 things that other founders should definitely replicate in their company?

Well, I think for a long time, you know, it's just, there's an extremely sort of wide ability for the company to just truth seek on things. Like I think one of the things of just, you know, when you think about just going back to systems building is that. It's not as easy as just like, oh, there's a fast way to just build a specific feature. To do the same thing in Notion is actually, it might be a bit slower than another company because I think we deliberate a lot about how that can impact the broader system. Again, like it's slow in the interim, but like what it unlocks in the long run is quite profound, right? And so the analogy that I think of is always Lego because it's like, okay, like every block has a way of connecting back to all the systems they've built. And so when they think of like a new system to introduce, it probably like takes a while to like, okay, well, does it fit to all the things we've built before? But what it enables is that this insane creativity in the world, the fact that the same bucket of building blocks, you could create a car and I can create a Barbie doll with the same blocks. Is quite amazing as a platform, right? And so I think founders can think of that as like the modularity, the systems nature of it's very applicable for product, but it's also very applicable for how you build a company. I would also say like, I think we're also evolving our values. Like we actually did a 3.0 version because there are certain things that, you know, we also need to evolve. Like it was good for a few years.

Yeah.

One of the recent ones we added was essentially leave the building, which is, I think as you become a larger company, it's harder for you to really sort of meet the customer. Right. And so you can easily, especially with AI, you can really fall into the trap of, wouldn't it be cool to build that? But it's like, but no, no, no. What are you trying to solve? Like, what exactly are the workflows that you're enabling? Just the fact that we can think of everything that we're doing with a customer in every room is something that becomes harder as you scale as a company or as an organization.

So let's shift gears towards this reinvention of the product that we've talked about in the AI era that you guys have really been at the forefront of. So you guys got access to, I think, GPT-4 or GPT-3 at the time, way before it was released. You were the first, one of the first companies to launch AI features.

Now, agents.

Was there a specific moment or a specific few weeks, months where you felt this is not about just adding AI features, this is about rethinking the entire product from the ground up?

Well, it's been a journey. I think it's like 3 and a half years since October 2022. We had early access and I think Simon was actually playing with it in the early days. And I think he was actually nudging both Ivan and I to pay more attention. But of course in 2022, it felt like, oh yeah, like, you know, everybody's been talking about AI for a decade now. Like, you know, I'm not sure if it's like, yes, yes, it's important, but there's also a lot of other things we need to do. But we were at a company offsite when we just had a little bit of downtime and we essentially just, you know, connected GPT-4. It's like, you're writing a doc, let's see if the GPT-4 just like completes the doc. Or maybe you highlight something. Let's see if it can just like, you know, you, can you tell it to just edit it? And it did. Like, it was like the, it was actually somewhat of a wrapper, but the fact that it was integrated into where you write and where you do your work felt quite magical. And if you saw that demo, like, I think it was quickly very clear that the way we do our work was going to change. And I would say the early days, it was like, I think the thing that kicked in was probably more defense. It's like, okay, well, there's just going to be a lot of companies who are going to try to be Notion, but with AI. And before that happens, let's just be that company so that like nobody can unseat us from that. So the initial sort of push was like, let's just defensively protect how people think about Notion today. And it turns out like that worked well and like became a business and it sort of moved into offense where it was meaningfully adding growth to the business, right? That's like meaningful percentage points added to the growth. But it's been a journey and I think it's gone from like, I don't know, 5 people built the first version to 40 to, I don't know, it's like 400 people who work in engineering all work on leveraging AI today. Yeah, we've learned a lot. Ivan talks about working with AI as like brewing beer. So we're on the 7th season of brewing and, you know, first couple seasons were okay. Like we got some people to notice that, but some of the releases last year really kicked it into shape.

Right.

And I feel like the business really inflected, the usage really inflected.

We got better at Even you've described the first couple of releases or the first couple of years of playing with AI as a swamp of despair. Were there some product mistakes you guys were making? Was it the limited capabilities of AI that was leading to the swamp of despair? Like what was it and what changed?

So the, yeah, so there's a writer piece, which I think sort of worked like plug and play. Then we did search so that you can sort of, you know, searching for pages, you can ask questions. And then after that, we built, we were trying to build the agent where essentially the idea was that instead of you having to manually do it, you can just say what you want to do and then Notion will just figure out a way to do that for you. That I think we had to build like 4 different times. And we basically hit a different blocker every time. The first time I think we were trying to do custom model with RL and it was just like the iteration speed was too slow. I think it would take 4 weeks for the model to live. It's like some things would work, but if this thing would work, something else would fail. So I think we sort of gave up on that. The second time we were like, oh, the model capability is much better. So, but I think we overshot ourselves in terms of the ambition of what we're trying to create. And so then we had to like shut down that one too. Then we decided, okay, like let's just start with very basic. I'm going to say, hey agent, put the word the on this page right here. Can you do that? Can you do that reliably? And we went from that to like, okay, do something more, something more. Like, and then eventually we were able to teach agent all of the APIs that we use ourselves as humans clicking around. And that release finally landed last fall, right? So it was just a few months ago. And so, yeah, it's an interesting new technology where of course the capability is changing, but also it's an interesting world of product development where just because it didn't work a year ago doesn't mean you can't, you shouldn't try again. Because you could try the same exact thing, but the capability is now ready for it to work. And then the other thing is like, and I think it's like a broader topic, but we have to, you have to re-architect the way your company works. You have to re-architect the way the tech works to make sure like you can flow with the model a little bit more and more.

I 100% understand your point that some capabilities are just the same thing you try again 6 months later, it might work this time, but Was it some aspect of the productization that worked better? Was it the combination of that with the models getting better or was the user more ready for it?

Like, I mean, I think if you think about the writer product, it was like a feature on top of the writing product. And because it was right there, I think a lot of people used it and paid for it. But then you take the second product, which was Q&A, the fact that you can't only just search for pages, but now you can ask for questions. I thought this was, this is probably one of the best products we ever built. And it required like a lot of infrastructure to vector databases and like make sure that we can do this at scale and offer it to every customer. Just in terms of like capability, I feel like enterprise search, what we built is like one of the best pieces of AI products. But it turns out that like human behavior is hard to change, right?

Mm-hmm.

So even though like you can ask a question to get an answer, nobody knows that you can do that. People still search for pages. And so then you have to figure out like, okay, well, how am I going to change people's behavior to even for them to even understand that they can do this? We tried to merge search and Q&A so that like maybe you search for something, but then you get an answer and maybe you're delighted by that and you will start doing more of that. But it was an interesting one where the technology is just in the product capabilities so far ahead than what people use it for. And so it took some time. Only after we connected to other tools was people realize, oh, this is different, right? So initially just having search become Q&A capability was there, but the usage was not there. Then we said, okay, not just Notion. Now you can connect your Slack and Google Drive and Outlook and Jira and Salesforce. Now people realize, oh, now I've connected all the different tools. I can now query all the data I have. And suddenly people realized, oh, this is like a different way of using this product, which started to click. So I think that that part is there. And then I think from like a sheer talent and team standpoint, you know, it went from a few people building the early feature to just that small nucleus sort of growing. And one thing, one of the things we realized was that whatever this group of 30, 40 people were just operating at a completely different velocity. than the rest of the company. And we would be puzzled because it was like these people were just always online, always on Twitter, always ahead of like all the models and they were just working a lot harder too. And then the rest of the company was in like the previous gear, right? And so, okay, how do you bring the rest of the company along? Instead of saying to this other group that, hey, you see this other team, like just do more of that. We decided the way to do this inside the organization is just to make this this small team bigger. So just like keep expanding this almost to a point where there's nobody left. And so at this point inside the company, you know, either you're in infrastructure and you're really thinking about security, reliability, scalability, or you're building product. And if you're building product, like there's nothing that's not AI. AI is just the fundamental technology.

Notion's product also has a lot of surface area. It's a vast product. So how do you guys balance the vastness of it with ease of use and educating the user that you can do— because again, you introduce a new AI feature, somebody like a Notion user might not even know, okay, they introduced this new AI feature that is pretty impactful, but in that corner of the product, which I don't otherwise touch.

Yeah, this is where you start to sort of like eat your company out from within a little bit, right? So Yeah, like it starts with a feature, but then you quickly realize, like, I think at least for us, we realized 2 years in that what did we sell for the last few years? We sold a system of record, right? We sold a great modern editor. We sold a great database, general purpose database product that you can use to build lots of different things. Those things were in the foreground. That's what I would come to you and be like, hey, Nakul, you should buy this. All of that stuff moved to the background and that's hard. For people who just like spent a lot of time building these things to be like, wait, now we're just putting this in the background? And we had to put it in the background because the human expectations changed. It used to be like the simplest unit of work used to be writing a note or a doc. Now the simplest unit of work is having a quick chat with your agent. The thing that Notion benefits from is like we've actually spent many years building an amazing system of record. It's extremely useful now.

Yeah.

But just the way people want to engage with it is through the agent, right? And so instead of people writing things inside a database or a doc, that is still what people will do, but they're just doing it through the agent more and more. So that's, yeah, that's profoundly changed how you build products. Like another example would be, you know, in the past, let's say if you're taking like inbox as a design, right? In the past, you would create a few different designs for how you want to change the flow of inbox. Maybe there's a different button, maybe there's a filter, maybe you collapse a lot of comments on the same page to one thing and so forth. And you would create these different design mocks and maybe you start building towards that. I think today the product manager and designer and the engineer, instead of doing that, I think what they have, like a better way to do it would be like, okay, well, just assume that people are going to try to manage their inbox by just telling the agent to do it, right? And so it's almost more important for them to understand how LLMs work, how agents work, how tools calls work, right? So it's like, I was like, don't do any design first. Just make sure all the notifications are easily, you can make easy tool calls and you can like do things with it. So just see what LLMs are able to do with your inbox. After you've figured out like, okay, this is how the LLMs can actually make the inbox better is when you get to design, right? It's almost like you have to rewire the way it looks and flows instead of clicking buttons to like, well, people just want to tell the agent, like, you know, archive everything that's coming from this person or whatever. And the LLM should be able to do that. And once you can prove that that happens on LLMs is when you get to designing the product.

Have you guys also then thought about acquiring AI product DNA versus Change is always tough, takes time and all of that versus infusing just AI DNA by acquiring small startups. I know you guys did Campsite, but also in the past you've acquired Calendar.

Yeah, actually this is one of the things I was looking at. You know, there's probably like, um, 20 or so Cannons or DRIs inside the company. I mean, these are people who can run, you know, sort of projects end to end and I was just trying to make a list of these 20 people that in my mind who I think I could just delegate fully on the product engineering side. And in my last count, I found 12 out of the 20 are either founders we acquired or founders we hired directly. And I think there's a sense of just resourcefulness and urgency that founders bring. That is quite special. I think whether it's AI founders, it's like, it's still early. Like, you know, I think if you've basically working with this AI, you've been working on it for the last 3 years. But I think if you're plugged in and you're sort of like on the bleeding edge, that's probably a good thing. I also think there's a new way of just building products that I think like we're evolving ourselves as a company, but also I think some of that we can sort of— inherit by acquiring a company or a founder.

Every company in today's world is also worried about OpenAI, Anthropic entering into their space. How does Notion think about that? Because like OpenAI has now, or at least at some point launched a meeting notes product and like, do you guys think about those guys at all or no?

We work quite closely with them, especially because we sort of leverage their models as well as other models in terms of building a lot of products we've built. Increasingly, we're more and more of an infrastructure product that allows humans and agents to work together. And sort of what we've built through the system of records, through like a multiplayer collaboration layer, through sort of like the permissions and all of that. allows more companies to be able to leverage AI at scale, right? And so I think at least today, if you're using ChatGPT and Claude, it's very much like a single person view. And I think what we've been able to work on for the last many years has been really sort of thinking ourselves as like the collaborative layer that exists. And so today I can create an agent and I can make you an editor and so we can collaborate to create the agent, and then we can tell it the whole company can now leverage that agent. You know, if you think about like what is allowing a software engineer to have 5 agents work in the background while they're sleeping is actually the infrastructure layer that GitHub has provided, right? It allows people to review the work. It allows you to sort of like go back to a certain point and like accept it. It allows multiple people to commit to the same code base. It's almost like all this infrastructure exists. And that's why that infrastructure plus the coding agents being really good allows software engineers to just get all of that leverage. We think of Notion as basically providing all that infrastructure for general knowledge work, right? It's like our hope is that the CFO can also do that tomorrow. Our hope is that a marketer can also do that. And what they need is basically that hard infrastructure. And so, you know, the combination of the platform we've built, plus the ability for us to connect to all the tools you use, plus the ability to provide you all the LLMs that are existing in the world today, that combination is, I feel like, very complementary to the LLMs.

There's also a broader concern around differentiation overall with the speed of code gen. New competitors can come up very quickly and catch up to your feature set pretty fast. And so There's a lot of worry about commoditization. Every subcategory has 15 different players. How does Notion think about competition, not from the labs, but the next startup that's just copying a lot of your features and catching up faster than ever before?

Yeah, I feel like depending on like which side of the bed I wake up, like I feel like I have a different— I mean, in some ways I feel like there is like a real fight for survival. And I think that fight for survival is what has allowed Notion to leave the SaaS boat and jump into the AI boat, right? I think if we didn't feel like it's like the urgency, I don't think we would've had, we would've crossed over. But I also feel like structurally, I think there will be, this is like one of the greatest opportunities where a new guard is going to emerge in the enterprise. For businesses, like the old guard is falling. And it's basically creating this incredible opportunity for new players to become the new guard. And there will be many winners in this, in this new world. In that world, I think Labs will become incredibly successful. And alongside, I think Notion can also be very successful. It's just a question of like, okay, what's the value we're providing? And so like, if you're a company and you're thinking about like, why should I get Notion? If you believe that agents are going to coexist with the way you work, Then you have to think of like, well, what's the platform where I'm going to build this workflows? What's going to give me the controls and what's going to give me the flexibility to use every model possible? And what's going to allow the infrastructure for me to review the work and collaborate with humans alongside agents? Some of that, I feel like it's like an amazing time. You know, another idea in my head is like, For a long time we thought like, oh, you know, like there's the Microsoft Office and then there's the Google Suite and maybe like Notion's the third generation of that. You know what the crazy thing is? Like today's like suite is not going to look anything like Office or Google Suite. Like it's going to look very different. In fact, the more I think about it, the more I feel like it might look more like iMessage. You know, it's just like you, you're going to have a bunch of agents. That are going to tell you all the things you need to do, whether it's like, oh, here's the important emails or here's how you can prep for the day. And here's all these different things. So that's the big opportunity is like, okay, can we become this work layer? And if we can crack that, I think it's a big opportunity.

It's the opportunity and the threat of a new entrant, because that is the opportunity for the new entrant also that Notion might have all these things that they have already built versus I can be the iMessage equivalent and hooks up to Notion and gets that data out into the iMessage agent.

Yeah. Yeah. I think there's actually both of those is like the new upstart. It's always been a threat to the business, which is like, okay, 2 kids in a garage. At least with the AI stuff, what I found is like great demos, but very poor reliability and security is like, you know, something that companies will want to care about. And then there's the other side, which is the labs and maybe some days you can feel like, oh, One company will rule the world. And like, why are we doing even what we're doing? And I just don't think that is going to be true. Like, I think there is going to be many winners, at least in the enterprise, maybe in the consumer, it's like maybe winner takes all. I think in the enterprise, there's going to be many winners and there's going to be many different ways that people are going to work. If anything, like we're seeing creative destruction play out right in front of our eyes. And it's like company that's bold to just take the flag and run for it may have a real opportunity to become a generational company.

What's the thing What's the thing that's the hardest about this product reinvention that people are underestimating? And what's the thing that is actually not as hard, but people think is hard? Like both sides might be true for anyone who's similar to Notion, a pre-AI company trying to do this.

Well, I think the people part is always the hardest, right? I think people today are just extremely anxious about what AI is going to do to their jobs. And whether they'll have job security, you know, all of that stuff. So that anxiety is very high right now. And it doesn't help that, you know, there's a lot of rhetoric around just, you can do a lot with like 2 people or whatever. And so one of the things that we've tried to do internally is really think about investing in our people becoming more, you know, sort of going up the curve of like AI literacy, right? It's kind of like in the software programming, you went from like tab as a way to autocomplete to agent to many agents. And so if you just apply that curve, it's like, okay, well, if you're working in marketing or finance, like, okay, are you going up that curve? Like what's the tab equivalent for you? What's the agent? What's the factory of agents for you? And we're investing in that because we benefit from being a smaller company. So it's like, okay, well, we can really invest in our people to become that factory owner. And we tell our people, it's like, we're doing this. As a way not to threaten you that if you don't do it, you're going to leave the company. It's more like an investment in you because you're going to become much more marketable. And even if you don't, even if you leave the company to go do something else, hopefully like you're creating value for yourself because everybody's looking for those factory owners to be there. And so the people part is always harder. I think for us, like, as we thought about just like the small nucleus we had and we tried to grow that, I think that approach worked well for us. 'Cause I think people who were outside of that circle wanted to be part of the circle. And then those people sort of bought in and decided to work on it. And then there was people who were just like skeptics or just did not believe in. And some of those people decided to sort of choose to leave and that's fine too. So I think that part is probably the people part is probably like just the way the macro is. I feel like that's probably the hard one. I think it's a little bit of like a way of operating has to change, especially around technology, I think. So just the example. like at highest level is like we tried to get the LLM to understand the Notion language. And we sort of kept hitting like our heads against the wall to like make the LLM do it. But instead, if you can change that to just like channel the model, which is like, okay, how do LLMs behave? Can we just like get on the same page? It's like, okay, they like Markdown, just give them Markdown, like all of that stuff. And then things start to happen. If you think about like, okay, well, there's companies that have been around for decades and like they have a certain architecture. of how the tech and the product works. If that's at odds with the way LLMs work, I think it's going to be harder. And if you can somehow re-architect the product you've built in a way that is more cohesive, more conducive to how LLMs work.

That's a very insightful thing you're saying. If I may double-click there, you're saying the LLM doesn't need to meet my product where it is. My product might need to meet the LLM where it is.

Yes. At least till the point where LLMs are good enough to do any architecture.

Yes.

Right. And so if you imagine like there's going to be a gap between like LLMs just being able to build anything and with anything, what if you're an incumbent or a legacy company or like, you know, you have, you've been around for many years, you really have to stop trying to fit the LLM with what you have and you have to fit what you have to the LLM.

Yeah.

And if you can somehow do that, You'll feel like you're going with the tailwind as opposed to the headwinds.

This part that you just said can also be extended to the people side of it, which is the people who are stuck that the LLM doesn't do this the way I do it, they'll keep rejecting it and they'll fall behind versus the people who themselves are malleable to meet the LLM where its capabilities are today and then evolve with that might just be on that AI train and suddenly Overnight, they'll feel like it'll be like an overnight AI success, but actually for 2 years they've been tinkering with the whole thing for a while.

I had breakfast with an engineer who was in the Tesla Autopilot team maybe a few months ago. And we were talking about open weight models and agents and he was saying that they're using it a lot. And initially I was a bit surprised, wait, like the Autopilot while I'm driving? You can use agents and open weight models. He said that, yes, like there's parts of the code where we're going to look at every line of code and make sure it's good and reliable, like, you know, secure and reliable. But there's just a bunch of different parts of the code base where we have just very strong eval set. If the evals get better, then you have, then we can ship that.

Right.

And so within those areas, I actually don't even care how the code looks as long as my evals just keep getting better, which to me was very insightful. Right. It's like, we just think of like, okay, well, everybody can code. So like, what's the review? Uh, is the new backlog, but actually you have to sort of think of it differently. It's like, okay, is there a different way of operating that is going to give you more leverage? And yeah, it's like, this is like this weird piece of technology that nobody understands. Like, why do LLMs behave the way they behave? Nobody can really sort of tell you that. And so you sort of almost have to like think of like, okay, well, how do I channel that model, channel that LLM more? So that I can get it to do all the things I want it to do.

So let's shift gears towards the people side of it since you've already touched that. First of all, do you believe that Notion is today a fully AI-native organization in every pocket of the organization like finance, sales, marketing, customer support, or do you feel like there are pockets that are still behind?

It's like all going on that curve. Nobody's on level 1. I think if there's probably like 1 to 5, I think some departments are probably at level 4, some at level 2. And interestingly, like one of the benefits of having some departments sort of further along is that it's sort of almost created like internal competitiveness. I think like actually engineering was probably much further along. And then our CFO decided, Rama decided to really lean into it. She told the entire org, like, there's no going back. We all have to do it in a different way. You have to use our existing tools, use other tools, but you have to rethink how we're, how we run finance. She called it Finance 2.0. And even in a couple of weeks, they made so much progress. Instead of taking 10 hours to close the books, they can now do it in half an hour, like, you know, by just by leveraging agents and seeing finance do it, like really got a lot of other departments to be like, well, what's our 2.0? So there's, you know, Legal 2.0. And so, so I think like people are sort of leaning into it, but there's still a lot to be done. And part of it is there's always, it will always be a work in progress because the goalposts keeps moving further ahead, right? It's like you can do even more things with it. I'm really excited about like go-to-market, like rewriting of that. It's actually kind of takes me back to my early days of Notion where I was trying to build a system that sort of like scales well. And then I feel like in the middle, we scaled it like how every company scales it. But I think we're coming back to the original ethos of like, okay, well, what if all of these things were more connected? Yeah.

Yeah.

What could that mean for where go-to-market teams evolve?

So if you don't mind, let's go one by one. So let's start with engineering. At this point, are your engineers writing any code at all or is it their role has evolved to reviewing?

I think most people are leveraging many agents together. I don't know if there's like a certain percentage, but probably high percentage of, you know, probably vast majority of the people are sort of like doing work through the agents. And then we're working a lot on just like making sure like the downstream effect of what that means is also Evolving to support that.

If you had to guess today or guesstimate today, how much of Notion's code, fresh new code written today is AI generated?

Oh, it's probably like, I don't know, 80% plus. Okay.

So 20% is still humans writing the code.

That would be my guess. I think there's probably like parts of product that does require us to be either careful or we still have people who are still like going up the curve. I think the combination probably makes me feel like there's probably still 10-20%. There's also some people who are just like much more familiar about certain parts of the code where they just feel like they can get to the answer faster. So like, yeah, my guess would be like 20% still.

And so has all of that then also created, you touched on this briefly, the surface area of buggy software, security issues, all kinds of downstream issues. How are you guys investing in the systems that need to catch that? Because the amount of code gen is probably 10x. From even a year ago.

It has. Yeah. I think it's a little bit of just like evolving the underlying systems. Um, so, you know, you can also use a lot of automated quality, uh, testing that we used to have humans do it before. Uh, it does require you to write more tests on this. Specifically around security, we just have to harden a few areas. You know, we use like a lot of other companies, a lot of different libraries, and we have to just be on top of, I don't know, in the last month there's been more leaks than in the previous year. Right. And so I think knock on wood, so far we've been okay, but it's like a constant battle.

But is there an ownership element of it? Like, have you created a crack team to say, hey, this is going to be a downstream issue, if not today, in 6 months. And if something goes wrong, who's the team accountable for that? Or is it each engineer still like each team?

I'm a fan of just having responsibility be, you know, with the people building as opposed to like a separate team that's going to like fix the issues. And so, yeah, I think there's been more SevOne incidents around like, you know, just because we depend on GitHub or we depend on another tool that is down. And so we have to sort of support that. There's also been other tools where there, there's like, you know, Just libraries that have leaks that we have to be on top of. I think like whatever the HackerOne submissions have gone up a lot as well. And so there's actually one that was pretty interesting where almost someone used an LLM to find something in Notion, then used LLM to submit it, the HackerOne bounty.

What a world to live in.

We're living in a world where agents are finding the bugs and then like we're using agents to fight it too. And so it's a very dynamic world right now and I think it'll just continue to evolve.

How has the head of engineering role evolved in this world?

Well, I think like Notion's always been a little bit more unique in this space, like where Notion's never been like a waterfall, like product manager to design to engineering. I think we've always been pretty blurry on the edges. It's almost like EPD is like a blob. So my engineering partner would attend my staff meetings. I would attend the engineering meetings. Out of the 20 DRIs I talked about, like half of them are engineering. It doesn't have to be a PM. And so I think if you think about like broadly what that means for engineering product design is that each person is able to do all 3 of those roles, or at least is able to lean into it more. That's always been sort of Notion's past as well. I think that's the reason we were able to sort of keep a very small product design team. How it's evolved, I think that engineering today is like, at least at Notion, is like these 2 parts to it. One is like, okay, well, how are we keeping the product in the frontier for what we need to do to just be relevant? And then the other part is like what we just talked about, infrastructure, security, reliability, scalability, and you sort of need those 2 vectors to balance each other out. The more you're in the frontier part, like the more you have to like balance it with it. There's a week where we push harder, like, you know, doing a lot more commits, a lot more new features, a lot more new products. And then we have to take a breather and make sure like the systems get hardened. And so as a head of engineering, you have to sort of balance that.

Sheer amount of visibility the head of engineering now needs to be aware, like the sheer amount of surface area that 10x code gen is leading to? Are they just like, yeah, overwhelmed with the amount of visibility they need to have on various parts of the engineering organization?

It's an interesting sort of dynamic world where there's like more stuff happening, but then there's also like more tools that give you leverage. It's kind of like, you know, like every model price is falling, but there's also like new models and new capabilities coming up. And so it's very hard to predict like sort of like where it's going to land. I feel it this way. It was like, you definitely feel a lot overwhelmed. It's going on just like the pace is just so much faster.

Insane. Yeah.

And so I feel like there's a lot of alpha in taking a breather and sort of, you know, calmly thinking about like, well, what are the bets you have to make? Like, especially if you're living off of like Twitter, you just feel like you just have to run like. every hour to a different direction. And so for us, it's like, I think we have a very clear product roadmap around like the infrastructure, like the hard things we're trying to build. To build all of that, we also need the ability to like scale for different customer sizes and do it, you know, in a secure way. And so it's almost like if I compare what we're focused on now versus like a year ago, we've become a lot more focused as a company, which is counterintuitive because you would feel like we should just be building more. And, but I think in some ways it's become more clear. what our place in the world is going to be. And as long as we just like focus on fewer things, but just doing them well, I think there's probably like, you know, a lot more alpha there.

Building has become so fast because of engineering velocity. What's the new bottleneck in building?

It has become fast. And a lot of people are saying like, you can now build, like there's infinite quality software, but where is this quality software? Like, you know, we don't see it yet. It's funny actually, like, you know, as a product manager of the past, like it's been fun to watch just like the whole circle of product manager loop in the last year. Like basically a year ago it was like, oh, PMs have no jobs left. And like, you know, no use because like designers can build and engineers can design, whatever. And it's like, oh wait, like everybody can build. So like now we need a little bit of like, somebody has to decide like what to build and somebody has to talk to the customer. It's like, oh, maybe the PM is back.

Yeah.

No, no, PM is the only thing left because Basically anybody can build now. So PM is like the best job. And so I feel like we're in like a little bit of a, we don't know where it's all going to land. I see the same loop with college graduates. Like a year ago, we were like, oh, like nobody's hiring college graduates and like job postings are going away for engineers. And like even the Stanford software engineer cannot get a job. And now like, again, the job postings are again going back up and people are realizing that, oh, there's just so much software to be built. I feel like we're in the heat of the moment. We don't know how these things will all play out. If anything, if truly anybody can build anything, then the alpha is going to be around picking the right thing to build and doing it with taste. And so we might be back to what was already true before.

Have your sales and marketing orgs also been completely AI-pilled? How has their day-to-day changed?

I think this is a work in progress. This is the area that I'm really excited about because I think if I look back at our own sort of go-to-market machinery, like a year or 2 years ago, we just used a lot of different systems for doing this work, right? It's like whatever Salesforce and Marketo and Zendesk and all these different things. There's also like a lot of people who specifically proficient at these systems, like someone who's like a Salesforce expert and Marketo expert, and you had to hire all of them, right? Because Oh, you want to make some changes to Salesforce? You got to get a Salesforce engineer. They need to call someone and whatever, you know, it was very rigid. Adding a field could take like a week. Painful. But that was the only way because that's how go-to-market runs, right? What's happened in the last year is actually quite incredible because I think in some ways more and more of these companies are being pushed to create APIs. And so that's one piece. The second piece is like the coding agents have gotten really good, right? Mm-hmm. And if you, if you just combine the two now, instead of sort of some Salesforce expert who understands the Salesforce system, uh, being the bottleneck, now you can put like your product engineers and be like, hey, here's the APIs. Here's the coding agents. Like, can we just use all of these APIs of all these different tools, connect them all into an internal system that can actually give us like the customer, whatever, 360 view of the customer. And then can we actually leverage the LLMs now to actually like figure out what we should be doing? Like, should there be an email going out to like nurture this customer versus? And so I think that's always a dream. There's always the dream to build like this go-to-market system that just sort of just tells you what the next best action should be. But it was so hard to build that because you just needed like all these, you know, you needed almost to build like a supply chain where these expert Who can help you with these specific tools. And then everybody would just blame data. Like, that's like, oh, I don't have the data. Like, I don't have the data attributed to this. I don't have, but now it's just the API that's just making it so easy to connect to this all. And so now there's just no excuse. The API's connected all. You can now actually build the next best action engine all internally that can connect to the products you're building, that can connect to your help center and what it's enabling in my view is it's going to be like an incredible customer journey. You just don't have to think about this being your self-serve business and this being your sales business and this being your marketing leads versus sales leads, whatever. You can just be like, okay, well, here's a customer. Here's an account we care about. For this account, a lot of people sign up by themselves. For this account, here's some leaders who also signed up who may be like someone we can engage through humans. And for this customer, we are building this incredible journey so that it doesn't feel like my product is sending an email and my sales account lead is sending an email. We have that view for them, right? To me, that's something that I'm really excited about for this year because it's almost like the way we built these factories for software engineering is very doable now on the go-to-market side.

Do you feel you've already seen yet the 10x SDR, the 10x sales rep who are just like using AI to automate all the grungy parts of their work and focus entirely on closing or pipeline generation in a way that they're actually 3, 4, 5x more productive?

It's definitely happening at like an individual level. So there's definitely people who are able to leverage that and like 3x your average person. But have we figured out a system that can like operationalize this for every person? We're very close where I feel like now we have the data and the systems and the LLMs and the agents to be able to do this. Again, it's like back to human behavior. Like we have to rewire how they work. And My hope is that a lot of people will lean into that and will want to work this way because it will mean that a lot of the busy work goes away and they can actually talk to like 5 times more customers. And then I think there'll be some people who will resist that. And I think that that's going to be an evolution we'll have to make.

Have you guys adopted tools like Decagon or others on the customer support side where it's completely AI enabled?

Yeah, we've been customers of Decagon for, I think, over 2 years now. So I think on the support side, it's already happening and I think the next evolution of that is not just that whatever 70, 80% of the resolution happens, but you can almost like integrate the Decagon agent into the product itself.

Yeah.

Right. And so like, if you are using Notion today and you like click on the agent, you as a customer may not know that this is not support, this is the agent, right? But you may ask like a billing question there. But now I feel like we're getting closer to a point where a Decagon agent will recognize it's a billing question and can Either try to solve it right there with you or can connect you to a human. And so I think we're getting closer to a point where there's no difference between product and customer support. It's just like one customer journey that we can create.

Do you feel marketing, you know, when all of this started, I thought marketing would be the perfect use case because literally marketing teams are generating copy, whether it's multimodal text or images. At the same time, there's also truth in AI slop. And like, has marketing been able to adopt AI in a way that they've Made their copy generation and actual marketing collateral much more productive?

I think we're getting closer. I think as a company, we care a lot about sort of the brand. And so I feel like if like a sloppy ad gets through, I feel like Ivan will be the first person to like see it and complain about it. So I think we've been more careful with that, but I think we're getting closer to the capabilities being there and like the data. connectivity being there to be able to show people the right thing. It's also interesting, like a lot of people think of like AI and automation and agents as a way to do what you were doing before, but now you can do it with less people. What I'm seeing is like, that may be true, but I think the best companies are just raising their ambitions. And if you raise your ambitions for all the different things you can do, it's less about efficiency and more about like, can each person be like way more productive? We don't feel like we need to like rush to do a layoff or something like that. It's more about like, okay, well, each person, can each salesperson be like 3 times more productive? Can each marketer be able to like run 5 times more experiments? Like, because if they had a factory, be able to like experiment more, I think they could be a lot more ambitious.

You guys also have, I read somewhere that Notion is about 1,200 employees, but 700 agents running internally. I don't know if the 700 came from you, is it an official number or not? How much of the AI automation that different parts outside of engineering is seeing at Notion is you guys building your own agents to make yourself more productive versus like we use, we mentioned Decagon, like buy versus build internally given Notion is such a building blocks are already there for agent capabilities.

Yeah, new custom agent products has been just a complete game changer for us internally. And because it's been used not just by engineers, but like literally every person In the company, like they can spin up an agent, give it the right context, give it the right job description, goes ahead and does it, the job. I think that number is now, I think like for 1,200 people, I think there's something like a couple thousand agents live inside Notion. I just saw one recently where a lawyer used Notion Workers to connect it to all the live patents. And it also is able to look at all the new products we're building and the agent just like keeps tabs on both our work and the patent work and just suggests ideas for what we can do in the world of patents. We've, you know, brought DealDesk as an agent. The people team uses it as a way to like answer their questions and so forth. So like people can actually spin up these agents quite a bit. And in fact, next week at our developer day, we're introducing the ability for you to bring external agents.

Wow.

And so you'll be able to build, bring Claude and Codex and Decagon and all of these agents. to just work directly inside Notion. And then the other way is also going to be possible where you can take the Notion agent and apply it to, you know, we have the SDK for you to take it to whatever tools you use. And so I'm actually pretty excited about that future. We definitely benefit a lot from being customer zero. It's almost like I can tell you with pretty high confidence what product release is going to work because in some ways we are a mid-sized company. And if it works well internally, there's a high likelihood that it sort of works really well for small and medium businesses. But we're also not very precious about having to use only the Notion agents. I think in this new way of working, you have to be very open-minded about leveraging whatever tools securely to automate and do your work. And there will be an ecosystem where I think Notion will be a player, but there'll be other players as well.

How has your recruiting changed in this world? Like, are you looking for a certain type of person now?

Yeah, I think some of the attributes are sort of like doubling down on some attributes more than others. I think for us, we definitely continue to look for system builders, but we're looking for more curious, more high agency people. We are looking for folks who are interested in sort of leaning into the future rather than sort of being more skeptics. Although you don't have to deal with that as much today, I think, as you did a year ago. The thing with engineering is actually quite interesting. Like I feel like our focus for the last, I remember maybe 18 months has been almost like a dumbbell shaped, which is like, we're very interested in early career starters who know no other way other than the AI native way. And we're interested in like experienced architects who can really help us scale the backend systems too. Right. It's like, we've been actually trying to like go lower levels and higher levels.

Yeah.

But it's also been a function of us being very heavy middle level, right? Like we were very late to having interns or career starters because we thought, you know, we don't have like the mentorship bandwidth. So we hired a lot of like mid-level people. So we've been trying to like balance that out a bit. Yeah, I think it's like you want people who are curious and are willing to change their mind on things because like nobody knows the answer.

Yeah.

You know, I may say something 6 months ago that I may say the exact opposite today.

But like, as an example, is the engineering interview evolved to a place where they wouldn't clear it if they're not using AI agents in their interview process? Like, have you put in AI fluency in the interview process itself?

We have. I think like the, the, just the coding exercise has evolved a lot. You know, it's gone from sort of you having to write code to Just sort of being able to just see, okay, well, how do you use these agents? Like, you know, it's less about the final answer and more about the approach and more about your understanding of the fundamentals. So I think it's evolved. I'm not sure if it's like been like a complete shift from what we were already looking for in the past, but maybe it's like small changes that make sense in the new world.

Taking a giant step back with all the progress you guys have already made on AI, both on product and people side. What still worries you about Notion's future? You and your co-founders, what keeps you up?

What still worries us? I guess as I said, like some of it is just, you know, it's like a fight for survival. Like somebody was saying, you know, a lot of people after reading the recent Colossus piece, like, wow, like Notion's like figured it out. Like it's, and I tell them it's like, yeah, like we might be like the 99th percentile in the software bucket, but we're maybe like the 50th percentile in the internet. Like there's a long way to go. In terms of how people operate, like the newest startups operate. And I think we have to continue to fight to be relevant today. And I think we have to continue to rethink what our strategy is like and sort of how in this new marketplace of products that people use, like why is Notion in the mix? I think that's, you know, somewhat makes the job interesting.

Yeah.

Like our growth rate. has accelerated every quarter for the last 4 or 5 quarters. In fact, our growth rate this past quarter we just ended is 50% higher than the growth rate that was there like, you know, a year or so ago. But again, it's like, it's not, we're not adding $10 billion of revenue like Anthropic every month. So like, I think we have to think about, it's almost like all the things we thought was the gold standard in the software era have been demolished. Has been demolished. There's just like a new metric. And you sort of have to think about like, okay, well, how do we do that sustainably in the long run? Like sort of, you know, so that we're building something durable, but also how we like continue to be relevant today in this era.

Okay. Quickfire round now. The founder story you admire most.

You know, there's a few stories recently that I've just gotten to know about public companies that went down by like 95% and then coming back up. The ones that come to mind are the Carvana story, the Robinhood story, and then most recently I sort of heard the AppLovin story. These companies, not only private, but in like public, went down by 95%. And then these founders were able to sort of take the company back up. It's quite remarkable. I mean, it's hard to do this in the private market, but to be able to do this in the public is just another level of resilience.

Yeah. A book you'd recommend every founder should read?

I would just read a lot of the letters that Berkshire Hathaway or even sort of Jeff Bezos at Amazon wrote. There's just such interesting insights. In fact, you can get a pretty cheap copy of all the letters compiled together on Amazon, I think, for Berkshire. And you can go back to the '60s and '70s and it was like a much smaller setup and just read some of the things. And I feel like it takes my mind away from the daily treadmill of Silicon Valley and into the temple of long-term compounding. Right. And so I recommend reading those. Yeah.

It's interesting you've mentioned those 2 because Berkshire obviously is very long-term, but Jeff Bezos has this 10,000-year clock. Have you heard of this? I don't know where he set it up, but it's literally a 10,000-year clock that he invested in for long-term thinking to encourage long-term thinking.

I love those projects. I think there should be more of those. It's a good antidote to 10x, 10x, 10x.

A founder you call when you're stuck on an operating matter?

So my old boss, Jeff Weiner, who's the CEO of LinkedIn, is someone I call once every 6, 12 months. He's been a mentor, almost a guardian to me in the professional world. If I'm in a bit of a bind, he's usually the first call I make.

Your favorite AI product other than Notion, ChatGPT, or Claude?

Well, this is actually the interesting one that people struggle to answer because it's like on one side our worlds have completely changed, but on the other side, it's still so early to feel like there's a— if I had to pick an AI product that I admire, I would probably pick Tesla. The Full Self-Driving has gone from, I don't know if I should trust this thing, to like, I can now go from Pacific Heights where we live to Menlo Park without looking up once.

Most overrated piece of AI advice around founder Twitter right now?

There's just a lot of demos that like are impressive, but just like I'm in this like incredible WhatsApp group of like the most prolific agent builders. And somebody was saying like, oh, this thing is amazing, but I have to like spend 90% of my time maintaining it, but I live for the 10% when it works. So I feel like that's just the technology is amazing, but for it to just work reliably is underrated.

a perspective on company building in the AI era that most people would disagree with you?

I think this whole idea that like one person will build a single person billion dollar company is, sounds very lonely to me. Like if the business is going so well, I think it's a lot of fun to build it with other people. And so I genuinely think in the long run, I think we should think of like the abundance mindset rather than like, I can do the same thing, but now with half the people or a 10th of the people. I also just think some of the best projects have been people from different backgrounds coming together, building something fun and unique. And I hope that continues to be true. Like I don't want to live in a world where there's like single person billion dollar companies.

The one superpower you have without which you wouldn't have all the success you have today.

We talked a little bit about it. I just don't care about like how many people report to me and the, you know, Ivan calls me like a stem cell. Like I go there and I sort of like build something, then I like get out of it. I like that. It also, I feel like it suits my personality. I get bored easily. So like, I like the ability to like work on new problems, solve them, and then work on the next big problem.

Okay. So closing question for the founder CEO listening to this, let's say they are $50 million of ARR, they're a pre-AI company, real product, real customers. You know, now they're at the risk of being a dinosaur. What's your guidance to them these days? Like what should they start implementing right now for the next 90 days and to do what you've been able to pull off at Notion?

I think the first thing is like the terminal value of these old businesses, like you should just assume they may go to zero. And so I think the way to think about it is you have to put yourself, like take yourself out of the company and out of the business you have today and think about if you had to restart it, like how would you do it? And then you have to do it. Like I think there's no other way. You have to disrupt yourself to where the business is going. I would not sit on the $50 million business. And sort of think about how to grow that incrementally. There's just like vast amount of creative destruction happening in a good way, like just new companies evolving and you can kind of do it to yourself before someone else does it to you.

Would you take a week off, just think through the product from ground up and the company from ground up almost?

I think that can help. I think about like, well, when would we build the AI product was because we were on a week-long offsite and it just got us away from the day-to-day treadmill for you to be able to think clearly. And so if you're just working whatever the new 996 with looking at Twitter every day, I think it can be hard to figure out very clearly like what exactly, how exactly is the world going to shape? What's the role that I and the company play in this new world? And are we on that direction or are we like incrementally growing? Another analogy that I heard internally was like, you know, we were playing the go-karting game and it was like, we were doing quite well because of the investment in the AI. Like we sort of moved from go-karting to Formula 1. We just have to like rewire how we drive in this race. And we're middle of the pack. Like we're not like crushing it like we were in go-karting. And I think we have a shot at the podium, but I think we have to rewire the race we're in now. Right. We cannot take the way we were go-karting and apply it to Formula 1. And so that requires, it starts with the founders. Like, can you rewire yourself? And then you have to Just aggressively, urgently get your team to sort of go at it as well. It's like drive to survive.

Yeah.

You have to survive.

Akshay, this was great. Thanks so much for coming on.

Thank you. Yeah, thank you for having me. All right.