We start a mantra in the company: we want to be a jazz band, not a marching band.
I love that.
It's like a self-reflecting value thing. It's auto-equilibrium with me. So, it's like, "Okay, I'm a jazz band person." So, we've been hiring more and more jazz band people. Also, AI happened in the past two and a half years, so more jazz band people can really shine during this moment.
Okay. So, we had manager mode, we have founder mode. There's this new mode. What should we call it?
Jazz mode, maybe.
That's not bad.
Hey everybody. Today we have Ivan Zhao, the co-founder and CEO of Notion. I like to think of him as the refounder. He started the company a long time ago and in 2015 he didn't have product-market fit. He and his co-founder moved to Kyoto of all places and kind of restarted the company and rebuilt it. After that, they got product-market fit and it really took off. He did the same thing in 2023 around generative AI in an offsite down in Cancun and literally restarted the company, and in my mind he's the canonical example of how a SaaS company can move and become an AI company. So that's super interesting. The second thing that's interesting is just hearing how he builds the company. He's very AI-native in the way he organizes the company, the way he compensates people, the way he builds today. And it rhymes a lot with some other CEOs that I've interviewed, like Jack Dorsey a couple weeks ago. I hope you like it. What do you think of my outfit, by the way?
I think your outfit... your pants could be looser.
Looser? Okay. On the bottom or just in general?
In general. I like your sneakers. I don't know what they are.
Okay. They're Italian, Premiata.
That's cool.
You're known as a very fashionable fellow. Where do you shop these days here in New York?
New York has a good store near here in Soho/Tribeca called La Garçonne. Not like a designer, but more like a European, easy-to-wear style.
I like the vibe.
Thank you.
I gotta up my game. Ivan, just channel some nice energy.
Okay. Hey, I want to talk about CEO stuff, not just fashion. I've been doing this Sequoia thing for a couple years and a lot has changed in the CEO role. Like the whole manager mode/founder mode thing happened, and now it's founder mode and some new mode seems to be happening. Let's just reflect back when Paul Graham's founder mode memo came out.
When was that? Since 2022, right? 2021.
No, no, no. It was later than that. It was like 2024. My reaction, by the way, was relief because I was a quirky kind of founder mode type of person, and everyone was always trying to talk me out of doing that. What was your reaction?
I didn't have too much of a reaction to be honest.
Nowadays, Jack Dorsey is kind of redesigning his company to put AI in the middle, and Brian Armstrong is doing the same thing. My sense is you're heading this way as well. Can you talk about your journey there, your company's journey? What's new? How are you thinking about the organization? How's it changing?
Yeah, my journey—I have a small detour off of founder mode, which is like trying to delegate, trying to grow. Notion has been very small as a company in the early days. When we had product-market fit in 2019/early 2020, we were sub-10 people. The company became profitable because we were small.
Did you raise venture in the early days or you bootstrapped it?
We raised venture.
Okay, that must have been tough.
It was tough. The first five years, just seed money dragged us for five years.
Yeah, HubSpot was like that. Took us a long time to get the product market fit.
It's very tough. But it's a different type of tough compared to now. Back then, there's no product-market fit. It's despair. You just black out, you don't know where to go. So we had been very small from the get-go and we sort of tasted the sweetness of it because you're small, you're profitable, it's easier to raise money, you're default alive. And we've been resisting growing team size. During COVID, a lot of companies really blew up, remember that?
I'm sure you did.
We relatively did, yeah, blew up.
You got to product-market fit at the right time.
We grew to product-market fit around the same time. There was a lot of advice that you've got to grow, so we did grow relative to other companies, but we were still resistant to growing too much in terms of headcount. But that was the moment I sort of learned, "Oh, you need to delegate, you need to introduce professional management."
Why? Who was telling you that?
It's a good question. Some investors... You didn't tell me that, I don't think so.
It was conventional wisdom amongst the investor class.
I would say there are pros and cons for that. We sort of resisted that for a long time. For example, we did not build a sales team. In retrospect, I should have started a sales team earlier. I thought I could first-principle the whole way, like creating a new kind of PLG, some kind of brand-new motion. That was a mistake. The other part is product marketing, engineering—those functions, I wouldn't call them a mistake, but we were upping the traditional way of building it.
Did you hire "been there, done that" people in that phase?
We did.
And do you still have any of them?
Many of them are still with us. Our CTO has been here for the last three years, and before that there was another CTO during that phase. But I would say some parts of Notion's DNA are defined by the classic SaaS executive.
Now, things have changed in the past three years since we've been building with AI. We were one of the earliest adopters of language models in our product, and you start to think things are breaking. Building with this new technology, the classic boundary between roles doesn't quite work. The analogy I like to use is that building classic software is like engineering a bridge—if you can design it, you usually can build it, it's fairly predictable. A product manager talks to customers, passes it to the designer, who passes it to the engineers. Building with a language model, back then and somewhat still, is like brewing beer.
You can't truly predict the underlying thing. Good analogy.
You cannot tell the yeast, "Hey guys, go towards that flavor profile more." You can't do that. You sort of have to throw your best people in there and see what the technology gives you. So it's not customer-first; it's more experiment with the technology first. Although in the last year or so the models got so much better, the spirit is still technology-first-driven development rather than customer-driven-first development. That really forced us to change how we build product. Designers are the most productive when designers, engineers, and product people all sit in the same bucket, work with evals, work with experiences, and figure it out.
Inside your company, has one of those personas kind of dominated and the others fallen aside, or do they all kind of do the same thing now?
Our DNA in the early days was always blurry. We were one of the first to hire designers who can code. I'm a designer who can code, so we built the company based on my image and my co-founder's image. Those profiles really shine: designers who can code, engineers who have strong product sense, and PMs who are not afraid to roll up their sleeves and try things. For example, one of our heaviest token consumption people in the company is a PM. Those people really shine, and we have quite a few of them at Notion.
Okay, so you read the Dorsey stuff, you read the Armstrong stuff. My summary of it is they think of it like a big triangle org chart. It's a very inefficient flow of information up, down, and across, it's been around for thousands of years, and it's time to rethink it. Stick AI in the middle, start delegating more and more over time, let the system make decisions, and feed it context in good ways. That's kind of my high-level summary of it. What's your take on what those two are doing, and are you doing the same thing?
In some sense, we're creating product for other companies to do that, but for ourselves, the high-level premise is similar. At least my interpretation is: knowledge work in the past was paper pushing. It required software to push information from one person to another, but so much of knowledge work is fuzzy, so humans had to be in the middle pushing information from one to the other. It turns out language models can do really well at two things: first, writing the software to design the perfect pipes for your company to push information and decisions; second, doing this kind of mini-decision-making in the steps between one decision point and another. A lot of people used to be in the middle of companies just pushing information and decisions around to sustain a business at scale. That can be completely rethought.
Okay. The analogy I like to use is that before steel, buildings could not be higher than five or six floors. New York City was largely six-floor-tall iron, stones, bricks. After steel, structures could go much higher. Language models plus software are the steel for organizations. What does the organization look like now that we have this steel? I know buildings got taller. What happens to organizations?
I think everybody's trying to figure it out. Dorsey's structure is like three layers or three types of people, right?
Yeah, three types of people that sit around the AI system that mostly train it, extremely flat. And Brian Armstrong said it's five layers and everyone's a player-coach, which I think is pretty interesting.
I think the spirit of all of that is right. I believe it should still be hierarchical. People have tried flat organizational structures in the past, but I don't think that's part of human nature. Human nature is hierarchical.
And do you think that's changed now that we have AI, that it can be flatter, or are you like, "That's BS"?
Oh, it definitely can be flatter. We are getting flatter and flatter.
Okay, how many direct reports do you have?
I have seven or eight.
And in general in the company, what is it?
Company-wide, probably around seven, eight, or nine people per manager.
A lot of people are going to 15, 20, 25.
Yeah, many of our best people have 15 or 20.
Got it. How have you changed your company? How you compensate people, how you build your org, how you hire people, just in the last year as all this stuff has developed?
Two years ago, we started to hire a different shape of engineers. We no longer optimize for people's experience; we optimize for people's agency, energy, optimism, and curiosity. A rough framework we are using is: Talent = Capability (experience) * Taste (value system) * Agency (will/drive). What language models changed is that, just like Google made information accessible to everyone, language models allow everyone to be a pretty good writer and programmer. So capability got normalized and democratized. Taste remains very important—your value system, what you want to bring to the world, which direction you want to go. You cannot change that easily. And agency—how hard you work—you cannot change easily. So we optimize for taste and agency today.
Okay.
That means in engineering, we hire much younger people, early out-of-school ICs.
This is the opposite of the way most of the CEOs I'm interviewing are going, which is interesting. They say a junior engineer powered by this stuff is more productive, but if you have a senior engineer who understands the whole system, has been there a while, and they are plugged in, they're like a 100x.
Yes. So we hire for a barbell shape: super junior and super senior.
Oh, interesting.
We hire for a barbell shape because the senior provides taste. There is still a lot of out-of-distribution information in architecture and language models that juniors are bad at, so seniors need to provide the direction. It's almost like a good engineer managing four to six coding agents at once, but a really good senior architectural engineer can manage two to three junior engineers, and if they each manage two or three agents, the leverage is higher and they can train the next generation. I think that might be a more optimal shape than a bunch of senior people managing coding agents directly.
Okay, and I assume you're still hiring designers and PMs. Is the ratio changing, or are you just hiring people to do the whole thing?
The definitions of designer and PM are changing. We intentionally hire designers who can work as PMs. Usually, it's harder for PMs to work as designers because the visual craft takes years to train.
But designers want to talk to customers and get feedback.
Yeah, it's rare, but don't stereotype designers; they do exist.
So if I look at your website, you're hiring junior and senior engineers. Are you hiring PMs and designers, or do you have some hybrid title? I don't know how to combine those two words together.
We hire PMs and designers. But when we interview them, we evaluate them with this hybrid lens. For example, we're in talks with two or three design leaders, and we evaluate them as PMs: "How independently can you drive things yourself? It's not just coming up with good ideas; how much are you interfacing with customers in addition to your design craft?"
Some of this stuff is like fashions changing. For the longest time, it was experience. Then the word everyone used was slope—"We hire for slope." And now it's taste, curiosity, and agency.
That's kind of slope, though. That's slope-ish.
How is it different?
Slope to me is just horsepower and smarts.
Well, intelligence has multiple dimensions, like GPU speed. Like, "How did you do on your freaking SATs?" But there are plenty of smart people who are lazy, that's for sure.
Do you call that high slope or low slope?
I don't know. I just always bucketed it as hiring people who learn fast and are very smart. And I think it's changed. The word "taste" comes up a lot, and you've been talking about taste for a long time, but now the industry talks about it. Why is that all of a sudden a thing? Why is that word everywhere now?
Because taste is not in language models yet. If you zoom out a long time, maybe it will be, but right now it's not. Taste is deeply rooted in your value system. Are you drinking sparkling water or still water? Diet Coke or iced tea? There is no right answer; it's to each their own.
Yeah.
The rest of the company, you've got more than just engineering. You've got a bunch of HubSpot people here, which is great, but let's just take marketing. Has the criteria changed there too?
We changed our marketing department. We no longer have a CMO organization. We did this recently—last year and this year.
Yep.
We broke our CMO organization into storytelling...
Yep.
...which is closer to product, because the product is changing so fast. Classic marketing can't keep up; we haven't shipped this fast before. So we have storytelling sit next to product, directly connected to social where discussions are happening. The other part is serving the sales go-to-market function, like demand gen and lead gen. We really focus on that. So rather than information making a round trip to a CMO who then figures out how to serve both product and go-to-market, we let both sides more decentralized figure it out themselves.
And are there any other classic marketing functions in there, or is it just those two?
Those are the two. On the storytelling side, there are also designers and creative folks, and community.
Yeah, community is in there. Community is part of our ecosystem, and startups are one of our most important communities. We have Notion fans around the world, and we figure out how to move our market through those people. That's the bucket of people.
And has the criteria changed for sales, marketing, and all the other functions? Have you changed the way you think about hiring? Most companies have a rubric when they interview someone—has that rubric changed?
We're in the middle of changing it. I just described the engineering and design ones that we changed two years ago. Marketing we changed about six months ago, and sales we are changing right now.
And when you hire a sales rep, do they have to be AI-enabled?
Yes. We changed our first interview: we no longer look at your resume. Build something for us, send a Notion link of something you built, and we look at what you built.
How about compensation? Is that changing? We had an environment where, at least at HubSpot, people were much more productive than the average person. Now it's even more so. Have you rethought compensation?
We are changing that as well. It needs to be much more of a meritocracy. You can't peanut-butter things. I think the SaaS era was kind of like a playbook peanut-buttering everybody; it was a pretty peaceful era. And right now it's wartime.
Totally.
It makes SaaS feel like, "What were we doing back then?"
Well, nothing changed for the longest time.
Nothing, right?
That's the reason I left HubSpot. It's like, okay, we're building this platform, we need to build more apps, we need to keep opening the platform, hire more reps, and just keep going. Until a couple of years ago, and now it's completely changed.
Or you can come back to be an operator if you like!
That's why I joined Sequoia. I was like, "Wow, there are so many exciting things going on." When we hit product-market fit, COVID happened. It was world-changing, but not in terms of how people build things and technology changing.
Do you prefer wartime or peacetime?
Oh, wartime is way more fun.
What's the difference? What do you do differently?
You feel more alive.
Okay. But what methods do you use that are different?
I need to take care of my body. I need to exercise every day.
I didn't think you were going to say that, but okay.
Yeah, it's just more fluid.
Are you more top-down versus bottom-up?
Notion right now is in a good spot with me in the sense that I can dance around really smoothly in either direction. There are enough people I trust deeply. They drive bottom-up, and it constantly surprises me. We are a company of a thousand people, and we have about 50 to 60 ex-founders in the company. They can really lead things. And they welcome me to jump top-down into any part I'm interested in or where I can provide value, and they don't feel territorial. The people who didn't feel that way are either no longer at Notion, or we let them go. So the company is in a pretty interesting equilibrium at the moment.
I interviewed Parker Conrad from Rippling, and he's similar—he has a lot of founders. The way it strikes me is that the bigger HubSpot got, almost all the important things cut across all those organizations, and it was just very hard to find someone who could really drive something across the org as opposed to just in their own silo. So we did not do that. I regret not doing more of that. Remember I said right after COVID we scaled, and then we tried to scale more traditionally? I realized that's not us. At that point, three years ago, we started a mantra in the company: "We want to be a jazz band, not a marching band."
I love that. The orchestra versus a jazz band—I like the marching band analogy.
Yeah, I realized I cannot be a marching band person. That's just not me. If I delegate everything and everybody does their thing, what do I do? It's a self-reflecting value thing. It's an auto-equilibrium for me: "Okay, I'm a jazz band person." So we've been hiring more and more jazz band people. Also, AI happened in the past two and a half years, so more jazz band people can really shine during this moment.
Okay. So we had manager mode, we have founder mode. There's this new mode. What should we call it?
Jazz mode, maybe.
That's not bad.
Because jazz mode has some structure, right? You want other people to contribute, you want people to have fun, not just you having fun or dictating everything.
Okay, let's give that a go. Let's push that out on X and see if it sticks. I mean, it's pretty good. I put a survey out the other day like, "What mode is this?" and asked if it was Dorsey mode, native mode, or other mode. Somebody put in there "death mode."
Death mode? I think that's a good idea. One other question on this new world: I work with all these startup founders, and planning is tricky. I sat through your pitch at Sequoia, and you had a plan and a P&L. What is it like internally, planning-wise?
Well, we have our financial plan, which most companies at this stage have if it works. I would say financial planning is useful to show how fast you're running, like you're on a treadmill: "Am I running at a seven, eight, or nine?" But product strategy? There is no freaking plan. No plan, because the market and the technology change so fast. It's not six months, it's not three months—it's week by week. So you have to jazz this thing. Financial things march along, and we try to be conservative to mid-conservative with financials and take risks elsewhere. And there's a buffer. Cost becomes another dimension—if you're just building software, you don't think about cost.
No.
Now, we really have to think about cost.
Can I ask you about that? A lot of these startups, their gross margins are awful. Companies like HubSpot had 86% gross margins—that's sort of SaaS normal. What is happening to your costs and your gross margins? And where do you think it settles out for companies like yours?
I don't have a final answer for you.
I'm assuming your gross margins have gotten worse, but maybe you didn't hire as much headcount, so operating margins are better. How willing are you to let gross margins get radically worse?
Oh, you have to. If you are not down for that, then you're not in war mode. I think Notion is for knowledge work, which might have a different curve than coding products. With coding products, you run frontier models—usually the smarter, the better. In knowledge work, we already start to see that you can settle with second-tier models, the open-weight ones. They don't need to be super smart. A lot of it is paper pushing information: "Oh, I spilled coffee on the carpet, can the HR team get someone to tidy this up next week?" That doesn't require Claude 3 Opus to file tickets. That's the kind of problem we are best at solving at Notion, so that will have a different gross margin profile than the biggest product-market fit today, which is coding agents.
Yes. Okay, you are in my mind the king of refoundings. You've done it twice—at least, Claude thinks you've done it twice. Your first one, you and your co-founder moved to Kyoto, locked yourselves in a room, and rebuilt.
It wasn't that small, but yeah.
Okay. More recently, of course, you went to Mexico, got early access to the models, and turned into an AI company. Talk about those refoundings and what your mindset was at the time. There are lots of early-stage founders who need to refound, and a lot of bigger SaaS companies that definitely need to refound.
Yeah. I like the idea that the best artists reinvent themselves. Miles Davis had at least three phases. Picasso had two to three phases. Michael Jordan's second phase probably wasn't as good, but Steve Jobs was the king of refounding. The first version of Notion took us four to five years to find product-market fit. In the middle of that, we found the shape of the product we wanted to build, but we were running out of money. So my co-founder, Simon, and I said, "Let's lay off everybody and go down to just the two of us." That's the Japan story. We rebuilt Notion from Japan.
How hard was laying off everyone?
It's not easy, but sometimes your body tells you that you have to do it. You're dead if you don't. There's a survival energy where you just do it.
Why the hell did you go to Kyoto of all places?
Well, we had never been to Japan. When you say goodbye to the people you worked with—Notion was not big, about five people—morale is so low. Simon and I wanted to go somewhere to change the mood, and we had never been to Japan. Initially, I looked at Tokyo on Airbnb but realized the houses were so small. I love Simon, but I didn't want to cram into a small apartment with him. Kyoto's apartments were much bigger because they didn't get bombed during World War II.
They weren't bombed.
They weren't bombed; they spared that city because of its treasures. The houses are much bigger and cheaper, so that's why we picked Kyoto. We rented out our place in SF and our office in SF, and we became cash-flow positive doing that. Once we landed, we just coded—coding, eating, coding, eating. It was really liberating.
I lived in Japan for a couple of years. Was there something about Japan that was inspirational in your refounding or the way you built the product? There's a certain aesthetic there that rhymes with Notion. Or could you have been anywhere—like Thailand?
Well, Simon likes warm climates where he can swim. Japan just happened to be a place we were both curious about. But once you go to Kyoto—you've probably been to Kyoto, it's a very special city.
Very special.
Yeah.
Very inspirational.
Very inspirational. Everything there is hundreds of years old. You open a door, and the shop is 150 years old. If you had gone to a beautiful place in Thailand, like Phuket, do you think you still would have pulled it off, or was there something magical about the location?
Most things are a story we tell ourselves, right?
Of course. Which story is the true story?
The story we tell ourselves is that Kyoto is a special place. If you can reborn in Kyoto, you can pull it off anywhere. It's inspiring to look at the surroundings—it's the craft capital of Asia. Notion fundamentally is a tool for humans, and those people are building knives, ceramic cups, tatami seats—they are building tools for other crafts. It's craft central. How can you not be inspired to build better software?
You still haven't answered my question: if you had gone to the Philippines and it was pretty, would you have pulled this off, or was there something exceptionally important about being immersed in such a craft-centric place?
I think I would have pulled it off, but the Kyoto story is a better story.
Okay, fine. Your big refounding was not easy, and you didn't immediately hit product-market fit afterward. It still wasn't easy.
No, it took a year and a half to rebuild from scratch.
Why didn't you just quit and start a new company at that point?
We never started a company just for the sake of starting a company. I got obsessed with this notion of a tool-shaped problem since my last year of college. Simon and I were both in this "tools for thought" community, whose lineage you can trace back to the computing pioneers of the Bay Area. I want to continue that lineage. To me, it wasn't like you had another idea kicking around. This is your life, and this idea is bigger than I can ever chew on. I wouldn't do anything else. Simon is in that community and obsessed with the idea too, so if we didn't work on this, we'd start another company doing the same thing.
Okay. I want to get back to the refounding, but just on the Grateful Dead—you studied important computer scientists from the 60s, and you're a student of Steve Jobs. There was a craft ethos about the Grateful Dead and the music back then, and certainly about Steve Jobs. Is that still here? It seems a lot more commercial than craft today in Silicon Valley. Have we lost that?
I think most people don't have it. Most companies don't have it. Steve Jobs, or people like him, represent the intersection of humanities and technology. Silicon Valley tech is largely a tinker culture. Tinker culture doesn't respect the science or history; people just like to build things in their garage, something works, they find product-market fit, and grow a business.
To be fair, Woz and Jobs were tinkering in the garage.
Yeah, but it required Steve Jobs to bring the human side of it. With tinker culture, you don't know what came before you. Science, at least, respects history. If you ask people in tech who Douglas Engelbart or Alan Kay are, most have no idea.
Honestly, the only reason I know them is I worked for Ray Ozzie.
Ray Ozzie, yes.
...who you would know because he's the father of Notes. He was obsessed with those two and talked about them constantly.
He was Microsoft's CTO for a while too. That lineage is broken today. The people who remember the 70s and 80s no longer work in tech. Tech doesn't know its past. If you don't know your past, you don't know history, which is humanity. Your view becomes only what's in front of you and what your competitor is doing. If you bring humanity, you have all other disciplines around you and history behind you. There's way more good stuff to steal from. I wish more tech founders stole from history rather than just copying what's in front of them.
I think they'd benefit from it too.
Back to the refounding: lots of founders are going sideways. Lots of companies that started in 2021 are going sideways, they've got 10 people, maybe they're cash-flow break-even. What's your advice to someone who has been going sideways for a long time, got venture dollars, and is working on product-market fit? Should everybody just rip the band-aid off, shrink it, move to Kyoto, and refound?
If their body tells them so. For me, you just feel that you have to do something drastic. During that time, Simon got his wisdom teeth taken out, so I was by myself. I remember being at my ex-girlfriend's studio and feeling this inner urge that there was no way out, I had to do this. And you feel liberated once you land in Japan.
I guess your advice would be that more people should listen to their bodies, be more risk-seeking, because they can get stuck in a venture-backed startup going sideways for 15 years.
There's no better time than now to do that because the market dynamic is wide open.
Yeah, I agree. Okay, you refounded a second time when you had 500 employees. That's a different animal. Without telling the same story you've told a thousand times, what was it like behind the scenes? What did you feel? Was it the same thing?
The second time was in Cancun, Mexico. We just got early access to the GPT-4 model. Our first interaction with GPT-3 was fine, but GPT-4 was a religious experience for me—a full-body religious experience. It was like, "Holy... we have to do something with this. This changes everything. Anything else we do will be meaningless." There were doubters in the company. Remember, three years ago was the crypto era, and people asked, "Is this another crypto thing?" Some of those doubters are no longer at the company, but the conviction was clear for me and Simon in the early days: we just had to do this.
Did you take two steps back in the business to go 10 steps forward? Did revenue slow while you were rebuilding everything?
Our first AI product had a meaningful revenue boost. We launched it two weeks before ChatGPT happened—an AI writing product. But after that, we wanted to build agent products in late 2022, and that took us a year and a half to really get working.
It didn't really work back then.
It didn't work at all. Simon and I might have been living too much in the future, which bites you with new technology. We tried everything: we built our own models, we worked with OpenAI fine-tuning, but none of it worked. That was a year and a half of no progress, and morale was low. Our revenue growth only inflected when the underlying models got better. But that was a painful period.
Was that the most painful period in Notion's history for you?
Different type of pain.
What was the most painful? Pre-product-market fit?
Pre-product-market fit is despair. You don't know where it's coming from, you don't know where to go. During the period where we were rebuilding our AI foundation multiple times, that was not great. Right now, everything is on fire, but it's a different type of pain.
Nice. A lot of 500-person SaaS companies are doing the AI thing, but they certainly haven't refounded. What's your advice to them?
Start with your product. Treat it like brewing beer, not building bridges, because at the end of the day, tech companies are about the product. The founder has to be really hands-on to experience that way of building.
Can non-founder-led companies pull it off?
I don't know. It would be tough. Founders have the moral high ground to change things, and people are more tolerant and forgiving. You have to use language modeling and feel the AI, feel the AGI.
Yeah. You can't just read about it or watch YouTube. You have to build it.
You have to build it. Building is the best way to feel it—build for your product, build for your internal systems, build for yourself. Tinker on the weekend, tinker on the side. Businesses are pathfinding entities in a market, finding local maxima for different niches. When a brand-new ingredient is introduced—in this case, language models—new paths open up. Each company has a different angle and niche. You have to internalize this new ingredient and figure out which path opened for you. If you don't know what it can do, you cannot find the new path.
I think you're right. You used a word that I cringe at, but is good: "calcify." A lot of companies calcify as they get older. If you're a company that's calcified, how do you get uncalcified?
We intentionally do a lot of acquisitions. We have 50 to 60 ex-founders at Notion.
And were they acqui-hires?
Acqui-hires, acquisitions. Founders are like little decalcifying machines trying to break things. They keep you regenerating.
I agree with that.
Yeah, it's really helpful.
Founders found for a reason, and now all of a sudden they're in a thousand-person company. How do you keep them motivated and not leaving to start something new?
First, some of them work on the domain they were working on before. The person leading our AI mail had an email startup before. The person leading our enterprise search is a founder of an enterprise search product. You give them a better platform and support. People care about the thing they are building. And you can argue it's more difficult to start a new company today compared to a year ago because the world is getting noisier, shipping cadences from existing companies are increasing, and everyone has Twitter fatigue.
And the SaaS companies got the memo and are shipping fast now. You were ahead of everyone, but HubSpot is cranking, too.
Yes, people have to kick into a different gear.
My view is that it's never been easier to start, but never been harder to scale. In every micro-segment, if there's any traction, you have a hundred competitors two days later. And in every vertical, there's a duopoly—a one and a two, and then fifty tied for third. When you look back in time, great companies were often founded in shitty times because people who are grinders and have a real vision start then. A time like 2026, you get a lot of tourist entrepreneurs with crypto-vibe energy.
I don't disagree. The market will do its thing; it's beautiful that way.
Advice to founders of reasonable-sized companies who are a little scared to refound—say, a CEO of a 500-person company where it's going pretty well, doing some AI stuff, and the board is happy, but there's a voice inside telling them they should refound. Does that person need to feel the AGI first?
If you feel it, you realize most of the tech industry supports different flavors of knowledge work, and we are just at the beginning of what this can do to the industry.
I agree. How do you think organizations look in three or four years? You're right at the center of this.
I think the truth is somewhere between Jack Dorsey's structure and traditional hierarchical ones. I like to look at the invariables. Human nature is static; we don't change that fast. As a species, we don't change, so there will still be hierarchy. If you watch nature programs, chimpanzees have hierarchies; it hasn't changed since we left Africa. Division of labor makes sense, and people have different personalities. There aren't 10,000 flavors of humans; there are probably a dozen, and they cluster around different preferences. Also, in our legal system, there are no autonomous companies yet—the CEO and CFO have to sign off. So it will start with a small team of humans. Collaborative systems must account for human nature. We need to figure out how this new ingredient, AI, can pass information and decisions around a group of humans.
I think of it as, right now, the AI system makes very few decisions—let's call it 1%. But as it gets more context, using Notion for example, it will make better decisions than humans. It'll have more context and history, and humans will spend more time training the system, giving it context and taste. The system will make more and more decisions, but humans remain necessary.
Yes. People say AI researchers in 15 months will not be able to make decisions to improve AI models themselves; it's the same thing, they're just on the bleeding edge.
By the way, you're incredibly well-positioned for all this. I worked with Ray Ozzie at a company called Groove Networks. We were in the knowledge management industry, which was a tough industry because Chief Knowledge Officers had no money and it was never urgent. Now, knowledge management is incredibly important and works incredibly well.
In the 90s, companies hired consultants to map out their org charts and information flows. That has now become AI deployment.
And your platform is ideally suited for it.
We forget that modern knowledge work is only about 150 years old. It was invented. It's not as old as fire or language. Why can't there be a new flavor of it?
Okay, I want to talk about you. I don't know you super well. You're soft-spoken, and I thought you were an introvert. Getting to know you better, was I wrong?
I think I'm still introverted. I prefer one-on-one conversations over one-to-many. I trained myself to do company all-hands by just doing them a lot.
A lot of founders are introverts, and the world is built for extroverts. Advice for founders who don't like talking to people or playing the extrovert game?
I had to evolve. I used to hate it, but to lead a group of humans, you need one-to-many communication, otherwise, people don't trust you. Writing is one mode, but face-to-face all-hands are even more important. For a while, I tried to delegate all-hands to my co-founders and other executives. There was no way around it; now, at every all-hands, I open the mic. I lead it. I'm pretty comfortable with it now. I still don't love it, but I don't dread it.
Advice for people doing all-hands?
Get a teleprompter.
I'm the opposite. I can't read fast enough for a teleprompter, I hate them.
English is my second language, so if I have to think and speak at the same time, it's hard. I prefer to have it written down beforehand. And now, local speech-to-text models are so good. Literally, the night before, I talk to myself, capture roughly what I want to say, and use a teleprompter during the all-hands.
Do you do them once a week, once a month? What's your cadence?
We are changing our cadence. It used to be once a month. Now, it's all-hands every other week, and an AMA on the alternating weeks, so we do something every week.
Do you ever step on yourself in an AMA? Do you ever say something you regret?
Yes, everyone does. Or you miss a good opportunity to rally people and give a non-answer instead. Non-answers are the worst.
At HubSpot, there would be a Slack channel during company meetings and AMAs. You'd go back and look at it, and if you said something stupid or missed an opportunity, it would really show up there.
That's good.
How do you manage your time, your schedule, and your deep work? How do you manage your energy? For me, my energy goes up and down during the day more than I'd like. Is that the case for you? How do you deal with that introverted vs. extroverted dynamic, and what's your schedule like?
It's not that crazy. I wake up at 7:00 AM and make coffee for my wife. I like to do work in the morning—do some thinking and writing at home—and then come to the office. I come to the office every day. A new habit I'm trying is going to the gym during the day.
That's a good idea, perks you up.
It elevates my energy in the middle of the day. I work in the office until around 7:00 PM, usually have dinner with my wife, and then do some more work until midnight.
When do you have time to go down the Claude rabbit hole?
That's my happy time! You are led by curiosity rather than responsibility. Usually on Saturday and Sunday mornings, I read books—sometimes work-related, sometimes just for guilty pleasure.
During the day, one of my founders here in New York has his schedule in 10-minute blocks. What does your day look like?
A lot of 25-minute blocks, with 5 minutes between meetings. My team packs my meetings together so I can have a one- or two-hour chunk of focus time, but most of my thinking time is in the morning or evening. I do a lot of Slack. My wife jokes that Slack is my social media.
Why doesn't Notion build a Slack killer?
We should, definitely.
A lot of your team said you're humble, and I think you are too. Why?
I wouldn't use the word humble. I'd use "truth-seeking." You seek truth, and you realize you only have one perspective. It's like the story of the blind men touching the elephant.
Yes.
We're just different blind men touching different parts of the elephant. Sometimes the tail is a snake, sometimes the trunk is a tree. That's my worldview.
Tonight we are going to have dinner with a bunch of founders, some with 10 employees who hope to build a 1,000-person company. What advice do you have for them?
It's a game—a social game. It has all the good and bad parts of a social game: humor, status, power. The CEO game is full of that, just like sports or high school lunch sessions. But it's also a pursuit of your personal values. I realized this in the past few years: I need to be in equilibrium with my own values for the company I want to build. That gives me sustainable energy. I can't just compete for competition's sake. I am very competitive and I want to win, but I need to be true to my values: building good tools, the craft, the human side of things. You need to figure out where you sit on that spectrum. If you only care about craft and not the business, you might as well just be an artist. It took me a while to find my equilibrium.
Are you someone who tries to fix your weaknesses, or do you lean into your strengths? What's your advice on that?
Definitely lean into your strengths. There are necessary things you have to do, like one-to-many communication as your company scales. Maybe you can lean more on writing than speaking. But within those boundaries, amplify what you're good at because capabilities are becoming commoditized by machines. Your unique perspective and strengths are what will last.
Something I assume isn't your strong suit is building an enterprise sales organization, which you are in the process of doing. I hear it's going exceptionally well. How did you get your head around it? How did you hire? It sounds like you hired the right person. So many PLG companies are desperate to move to enterprise, and the founders are in a similar mold to you. Can you talk about that?
We made mistakes. As I mentioned, we resisted traditional sales for two years. We wanted to first-principle our way into it and design our own system. That doesn't work. It's human nature—you want to talk to a seller. The modern sales playbook has been around for two decades for a reason.
I think enterprise sales is the last function...
I agree. You want to see a human. You don't want to get a diagnosis from a robot, so why buy an expensive product from a robot? It's not that hard, but smart people think they can invent something completely new. Fundamentally, you shouldn't reinvent the wheel unless absolutely necessary. I was trying to reinvent a new go-to-market motion, which was stupid.
Well, you built HubSpot.
Yeah, but inbound was the one innovation point that made HubSpot. Each company should preserve its innovation points for only a few places. You can't spread them everywhere, or you spread too thin. We tried to innovate on sales, but we should have just gone with classic sales because it works, though we make it more efficient internally with our own tools today.
Okay. So you woke up one day and realized you had to get good at enterprise sales. What did you do? What are the mistakes everyone makes? Did you miss a hire or two before getting the right person?
I thought this was solved today, but people struggle.
Enterprise sales has changed very little since I did it in the 1990s.
Initially, we tried a half-PLG flavor of sales with a systems-thinker sales leader, but it didn't work. You're just taking orders. PLG gives you the luxury of customers wanting to buy your product, so it's just fulfilling demand. You can't get sellers to do something difficult when there's an easier option. Then we realized we needed to respect the system. Our CRO is Erica, who was the CRO of GitHub—she's a systems thinker. We paired Erica with our head of sales, PVesh, who is the...
...meat eater.
Meat eater, right. Beans in his teeth. It's a perfect pairing. We have sellers with raw grit, but they respect working within Erica's overall system. This is when things started to work over the past year and a half.
Got it. In many organizations I deal with, the engineers look at how much money sales makes and they roll their eyes and complain. I assume that happened here.
No, Notion culture is very receptive to the idea that different people are different. We never had organ rejection of sellers. Our engineering team loves working with the sales team.
Okay, awesome. Speaking of culture, I saw online some people refer to you as a cult leader, and some referred to me as a cult leader back in the day. I didn't particularly like that. How does that land with you?
I like that.
Okay.
A company is sort of like a religion. You have a point of view and a value system that you project to the world through a product and business. My favorite saying is that the Catholic Church is one of the most successful companies of all time—2,000 years. They have their rituals and business models, and they're still around. They had a great founder in Jesus and a great head of sales in Paul, right?
Yeah.
And there are a lot of sellers, a referral system, viral marketing...
...viral marketing, OG viral marketing.
Yes. To me, a belief system or value system is one of the few things ingrained in human nature. People want to believe in something to find meaning and purpose. That's why wartime makes it more fun—your purpose and meaning are amplified, and the company gives people that camaraderie. It's the same reason we play sports.
The first four letters in culture are c-u-l-t.
Yeah, I should look up the root. What is an all-hands? People go to church every Sunday, all-hands every week—it's the same thing.
We used to have a goal with our all-hands at HubSpot where Dharmesh and I wanted it to be so inspirational that people would well up or cry. We really worked on them.
Yeah, I don't think there's anything wrong with that.
No.
Okay. Thank you for coming on the pod. Thank you for building Notion, and thank you for being an inspiration for tons of founders.
I'm glad, and hopefully, it's helpful for others.
Appreciate you.
Thank you for your time.
Okay, I hope you liked that episode with Ivan. One of the things I like about him—there's a famous quote by Oscar Wilde: "Be yourself; everyone else is taken." I like to think I am myself. I'm pretty quirky and I've embraced that quirkiness. And I think at the end of the day, companies really reflect those CEOs. And I don't think CEOs should be afraid to really be themselves and kind of build the company around them. Ivan's really done that nicely. That might be a good plan for you, too. The second kind of takeaway I had is, I work with lots of startups and there are so many companies, let's say less than 500 million in revenue growing like 30%, that are kind of stuck, and there's infinite numbers of stuck companies out there. If you're in that boat, I'd recommend a refounding moment. And the refounding is harder than it looks. It typically involves like a big step back and two steps forward. It's going to be hard. Your numbers will look worse in the short term. Ivan didn't do this, but when we wanted to do a refounding at HubSpot at one point, we started a new company inside of HubSpot that was completely independent from it, and that served us well. That's how Apple did it back in the day. But if you're stuck, and a lot of companies are stuck, be risk-seeking and do what Ivan did. Like, he was a 10-year-old SaaS company, and now he's a rocket ship AI company. I think he's onto it. The last takeaway I have on Ivan kind of starts back in a podcast interview I did with Jack Dorsey a few weeks ago. And it does seem to me there's a new way to build and run companies today. That's a pretty big departure from the way people have done it the last 30 years. Jack's iterating on it. Ivan's iterating on it. A lot of the call it vintage 2025 startups are iterating on it, and this is something I'm thinking a lot about. I think there needs to be a new playbook on how to be a CEO in the AI-native era we're in. And I'm early in the iteration, but just roughly the way I kind of think about the way they're organized is: historically, there is a triangle-shaped org chart. Those org charts have been around a long time, since the Roman Empire. They're a very inefficient way to run a company, but they've been the best way to run a company because there's been no better one. I think the problem with them is the bigger that org chart gets, the more that information gets distorted across and up and down, and the slower it gets. And so it's pretty inefficient. The new way that Dorsey described, and I suspect Ivan's is similar, is it's a circle, and in the middle is all of the context in the AI system. Initially, humans are making all the decisions, but over time, as you give that kind of centralized AI system more context, it gets smarter, and you're feeding in context and you turn over more and more of the decisions to it. I don't think it happens overnight. I think it's gradual, but it's an IT task to get it right. You have to be legible. I don't love that word, but you have to have a legible set of systems. You have to kind of record everything. But I think this is the new way. Companies who are doing this seem to be just moving a lot faster than the companies that aren't. And Ivan is moving at breakneck speed. And I think there's a lot of things that fall out of it: they don't need as many people, they move much faster because of that, they have much shorter planning cycles. All of those kind of "one-way doors" that Bezos used to talk about are really two-way doors because nothing takes that long to do. Anyway, I think there's this new playbook. I'm interested in it. I'm going to be talking about it in the podcast and on X. If you're interested in it, too, hit me up @Halligan. Thanks for tuning in.