Notion’s Ivan Zhao: The Refounder
In a Nutshell
Ivan Zhao refounded Notion twice—first in Kyoto after missing product-market fit, then in Cancun to rebuild it as an AI-native company. He is shifting the organization toward a flatter, "jazz band" structure where AI handles information flow and routine decisions, while humans focus on taste, direction, and high-agency work. Hiring now prioritizes curiosity, agency, and value alignment over experience, with roles blurring as designers, PMs, and engineers work directly with models in a technology-first process.
These notes were generated by AI and may contain inaccuracies.
Ivan Zhao introduced a company mantra: wanting to be a jazz band, not a marching band. This reflects a self-reflecting value system and auto-equilibrium with his own personality. He has been hiring more jazz band people, and AI developments over the past two to two and a half years have allowed these types of people to shine more. The conversation explores whether this represents a new mode beyond founder mode and manager mode, with "Jasmo" suggested as a possible name.
Ivan Zhao is described as the refounder of Notion. He started the company years ago, but in 2015 lacked product-market fit. He and his co-founder moved to Kyoto and restarted the company, rebuilding it from scratch. After achieving product-market fit, the company took off. In 2023, he repeated this refounding approach around generative AI during an offsite in Cancun, restarting the company again. He is viewed as the canonical example of how a SaaS company can transition into an AI company. Zhao is very AI-native in how he organizes the company, compensates people, and builds today.
Zhao took a detour from traditional founder mode thinking by trying to delegate and grow the company. Notion remained very small in its early days, with under 10 people around product-market fit in early 2019 to 2020. The company became profitable because it was small. They raised venture capital in the early days, which was tough—the first five years involved scraping for money. This contrasted with the current era, where the lack of product-market fit created despair and uncertainty about direction. The company stayed small and tasted the benefits: profitability, easier fundraising, and being default alive. They resisted growing team size significantly.
During COVID, many companies expanded headcount dramatically, but Notion remained relatively resistant. At the right time around product-market fit, they did grow relative to others, though still resisting large headcount increases. Zhao learned he needed to delegate and introduce professional management.
Investors and conventional wisdom from the investor class pushed for professional management. Notion resisted this for a long time. In retrospect, Zhao believes he should have started a sales team earlier. He attempted to first-principles the entire go-to-market motion and create a new PLG approach, which was a mistake. Product marketing and engineering functions were built in a non-traditional way.
They did hire experienced "been there, done that" people during that phase, and many of them remain, including the current CTO who joined three years ago. Some parts of Notion's DNA were defined by classic SaaS executives.
The past three years of building with AI have changed things fundamentally. Notion was an early adopter of language models in products. Building with this new technology breaks classic role boundaries. Zhao uses the analogy of building classic software as engineering a bridge—design it and you can build it predictably, with product managers talking to customers, then designers, then engineers. Building with language models is more like brewing beer: you cannot truly predict outcomes or tell the yeast to go toward a specific flavor profile. You throw your best people in to see what the model and technology produce.
This shifts from customer-first to technology-first driven development. The spirit remains technology-driven rather than customer-driven, though models have improved significantly. This forced changes in product building: designers, engineers, and product people sit in the same bucket, working with evaluations and experiences together.
Notion's early DNA featured blurry roles. They were among the first to hire designers who could code. Zhao himself is a designer who can code. The company was built in his and his co-founder's image, so profiles like designers who code, engineers with strong product sense, and PMs unafraid to roll up their sleeves thrive. One of their heaviest token consumers is a PM.
Zhao read about Jack Dorsey and Brian Armstrong's approaches. His summary: traditional organizational structures are inefficient for information flow, and the solution is sticking AI in the middle, delegating more, letting systems make decisions, and feeding context effectively. Zhao's take is similar: much knowledge work involves paper pushing that requires humans to move information between people. Language models excel at two things—writing software to design perfect information pipes for companies, and making mini-decisions between decision points. Companies used to need many people in the middle pushing information; this could be rethought.
Zhao uses an analogy: before steel, buildings could not exceed five or six floors, and New York was largely six-floor iron, stone, and brick structures. After steel, buildings could go much higher. Language models plus software represent the steel for organizations. When asked what organizations look like with this new capability, Zhao noted that Dorsey's structure has three layers or three types of people sitting around an AI system that mostly trains it, creating an extremely flat organization. Brian Armstrong described five layers with everyone as a player-coach.
Zhao believes the spirit of flatter organizations is right, but does not believe organizations should be purely hierarchical. Flat organizational structures have been tried before and do not align with human nature, which tends toward hierarchy. However, with AI, organizations can definitely become flatter. Notion is getting flatter and flatter. Zhao has seven direct reports, and company-wide the ratio is around seven to nine people per manager. Many of their best people manage 15 to 20 people.
Two years ago, Notion began hiring different shapes of engineers. They no longer optimize for experience but for agency, energy, optimism, and curiosity. A rough framework: talent equals capability times experience, times taste or value system, times agency or will. Language models have democratized capability—everyone can access information and become a pretty good writer and programmer. Taste and value system become more important because they are harder to change, as is agency or work ethic.
This means hiring much younger engineers, referred to internally as IC early out of school. This creates a barbell shape: super junior and super senior engineers. Seniors provide taste and direction because there is still a lot of auto-distribution information in language model architecture that no one is very good at yet. A good senior architectural engineer can manage two to three junior engineers or interns, each of whom manages two to three others. This may be more optimal than having many seniors manage coding agents directly.
Notion still hires designers and PMs. The definition of these roles has changed. They intentionally hire designers who can work as PMs. It is harder for PMs to work as designers because visual craft takes years to train. Designers are encouraged to talk to customers and get feedback, though this is rare. When interviewing PMs and designers, they evaluate how well candidates can drive things themselves, interface with customers, and demonstrate craft.
Hiring criteria have shifted from experience to slope, then to taste, curiosity, and agency. Slope refers to horsepower and intelligence across multiple dimensions. Taste is now embedded in language models and is rooted in value systems—personal preferences like choosing sparkling water versus still water or diet coke versus iced tea.
Notion eliminated the CMO organization this year. They broke it into storytelling, which sits closer to product because the product changes so fast that classic marketing cannot keep up. Storytelling connects directly to social where discussions happen. The other part serves sales and go-to-market functions like demand generation and lead generation. This decentralizes functions rather than routing everything through a CMO.
Community is included in the storytelling bucket. Notion fans around the world form one of their most important communities and help move the market.
Hiring criteria for engineers and designers changed two years ago. Go-to-market and marketing criteria changed about six months ago. Sales criteria are changing now. When hiring sales reps, they must be AI-pilots. The first interview no longer looks at resumes; candidates must build something and send a Notion link or artifact for evaluation.
Compensation is shifting toward more meritocracy. The SaaS era felt like a peaceful playbook where everyone received similar treatment. Now it feels like wartime. Zhao prefers wartime because it feels more alive and exciting. Methods differ: he exercises every day to take care of his body. The company has reached a good equilibrium where Zhao can move fluidly between top-down and bottom-up approaches. There are enough trusted people who can lead from the bottom up and constantly surprise him.
Notion has about 1,000 people, including 50 to 60 ex-founders. These people can lead effectively, and they welcome Zhao jumping into areas where he can provide new value without feeling territorial. People who did not share this mindset either left or were let go.
Zhao realized he cannot be a marching band person. Delegating everything would leave him feeling purposeless. This is a self-reflecting value that creates auto-equilibrium. He hires more jazz band people, and AI developments allow these individuals to shine. Jazz mode has structure while allowing others to contribute and have fun, rather than one person dictating everything.
Financial planning remains useful as a treadmill metric to know how fast you are running. Product strategy has no plan because market and technology change week by week, requiring a jazz approach. Financial planning stays mid-to-conservative with buffers, while product takes risks. Cost has become another critical dimension—previously irrelevant for software, now essential to consider.
Notion is willing to let gross margins get radically worse. For knowledge work, they can settle with second-tier models like Chinese models or open-source ones rather than frontier models. Tasks like filing tickets for spilled coffee do not require the most advanced models. This creates a different gross margin profile than coding agent products, which benefit from the smartest frontier models.
Zhao is considered the king of refoundings. The first occurred when he and his co-founder moved to Kyoto, locked themselves in a room, and rebuilt the company. The second happened more recently in Mexico with early access to models, transforming Notion into an AI company.
Many early-stage founders need to refound their companies, and many larger SaaS companies also need to refound. The best artists reinvent themselves. Miles Davis had at least three phases. Picasso had two to three phases. Steve Jobs is considered the king of refounding. The first version of Notion took four to five years to find product-market fit. During that period, the team discovered the shape of the product they wanted to build but ran out of money building it.
Ivan and his co-founder Simon decided to lay off everyone and continue with just the two of them. They rebuilt Notion from Japan. Laying off the team was not easy, but sometimes your body tells you that you have to do it. Ivan described it as a "panda energy" — you know you're dead if you don't do it.
They chose Kyoto because they had never been to Japan and wanted to change the mood after saying goodbye to their team. Initially, Ivan picked Tokyo on Airbnb, but the houses were too small. Kyoto apartments were much bigger because the city wasn't bombed during World War II, and the houses were cheaper. They also ran their place and office in San Francisco remotely, becoming cash flow positive for the first time.
Once they landed in Japan, they started coding in a simple routine: coding, eating, coding, eating. It was actually really liberating.
Ivan and Simon like warm climates, and Simon ideally likes to swim. Japan happened to be a place they were both curious about. Kyoto is described as a very special and inspirational city. Everything there has history — shops that are 150 years old, shrines. The city is the craft capital of Asia, with people building knives, ceramic cups, and tatami seats. This craft culture inspired them to build better software tools.
Ivan noted that most things are stories we tell ourselves. The story of Kyoto being a special place made the refounding more inspiring. When asked if they could have succeeded in a place like the Philippines, Ivan said they probably would have pulled it off, but the story would have been less compelling.
After the Japan refounding, it took another year and a half to rebuild from scratch and hit product-market fit. They didn't quit and start a new company because they never started a company just for the sake of starting one. Ivan had been obsessed with the "Notion-shaped tool" problem since college. Both he and Simon were part of the tools for thought community, which traces its lineage back to the first-generation computing pioneers in the Bay Area, including the Grateful Dead era.
Ivan wanted to continue that lineage of computing pioneers. To him, this wasn't just another idea — it was his life and bigger than he could chew on. If they didn't work on Notion, they would start another company doing the same thing.
Ivan was asked whether the craft ethos of the Grateful Dead era and Steve Jobs still exists in Silicon Valley today. He believes most people and most companies don't have it. Steve Jobs represented the intersection of humanity with technology. Silicon Valley is largely a technology and tinker culture that doesn't know its own history. Tinker culture doesn't respect history the way science does. Most people in tech have no idea who Douglas Engelbart or Alan Kay are.
The people who remembered the 70s and 80s computing pioneers are no longer working in tech today. Tech as an industry doesn't know its past. If you don't know the past, you don't know history, which is humanity. Ivan's point of view is that there's way more good stuff to steal from history and other disciplines. He wishes more tech founders would steal from the past rather than just looking at competitors.
Many founders are going sideways, and many companies started in 2021 are stuck with around 10 people, maybe cash flow break-even. Ivan's advice is that if your body tells you to do something drastic, you should listen. There's no better time than now to do that because the market dynamics are wide open. Founders can get stuck for a long time in venture-backed startups and spend 15 years of their career on something going sideways.
The second refounding happened when Notion had around 500 employees. It took place in Cancun, Mexico. They had just received early access to the GPT-4 model. Ivan described his first interaction with GPT-4 as a "full body religious experience" — he felt they had to do something with it because it would change everything.
Around three years ago during the crypto era, many people inside the company doubted whether AI was another crypto-like hype cycle. Some of those doubters are no longer at the company. However, Ivan and Sam had clear conviction that they had to pursue AI.
Notion's first AI product launched two weeks before ChatGPT and saw a meaningful revenue boost. They then tried to build an agent product at the end of 2022, which took a year and a half. It didn't work at all. They tried everything — building models, OpenAI fine-tuning — but nothing worked. Morale was low during this period. Revenue growth only inflected once the underlying technology got better. Ivan described this as a painful period but noted that the early pre-product-market-fit days were despair — you don't know where the pain is coming from or where to go.
For the many 500-person SaaS companies doing AI work without refounding, Ivan advises starting with the product. Treat it like brewing beer, not building bridges. The founder has to be really hands-on to experience that way of building. Non-founder-led companies would find it tough. Founders have the moral high ground to change things, and people are more tolerant with founders.
The person leading the effort needs to use language modeling and know it — you have to feel the AGI, feel the AI. You can't read about it or watch YouTube; you have to build it. Building is one of the best ways to feel it, whether for your product or internal systems. Tinker on the weekend or on the side. At the end of the day, businesses are path-finding entities. When a new ingredient like language models is introduced, doors open. You have to internalize this new ingredient to find which new path opened for you.
Ivan was asked how to uncalcify a company. Notion intentionally does a lot of acquisitions. They have 50 to 60 founders at Notion through acquihires. Founders are like little decalcified machinery trying to break things. Keep regenerating.
Some acquired founders work on domains they were working on before. The person leading Notion's AI notes had a startup in AI notes before. The person leading enterprise search is a founder of an enterprise search product. Giving them a better platform and support makes them feel appreciated and gives them leverage. They care about what they're building rather than fighting for product-market fit in a crowded market.
It's never been easier to start a company, but it's never been harder to scale. In every micro-segment, if there's traction, there are a hundred competitors two days later. In almost every vertical, there's a duopoly — a number one and number two, then 50 tied for third. The great companies were often founded in shitty times because people who are real grinders with vision didn't start in terrible times like 2022-2023. Times like 2026 attract tourist entrepreneurs.
Ivan believes organizations will still be hierarchical because human nature is relatively static. Division of labor makes sense, and people have different interests and cluster around a dozen flavors of personality. Legal systems require humans to sign documents, so there will always be a small team of humans at the start. The key question is how to use the new technology ingredient to pass information and decisions around small to midsize groups of humans.
Currently, AI systems make very few decisions in companies — around 1%. As AI gets more context, it will make better decisions than humans in some areas. Humans will train the system and give it context and taste. Over time, the system will make more decisions, but humans remain necessary.
Modern knowledge work is only about 150 years old. Knowledge management was once considered a joke industry in the 90s — companies hired consultants to map org charts and information flow. Now it's incredibly important and works well. Notion is well positioned for this.
Ivan considers himself introverted. He prefers one-on-one conversations over one-to-many. He trained himself to do company all-hands by doing them repeatedly. To lead a group of humans, you need to do one-to-many communications, otherwise people don't trust you. He used to delegate all-hands to co-founders, but realized he had to lead them himself. Now he's comfortable with it, though he still doesn't love it.
Ivan uses a teleprompter for all-hands because English is his second language and it's hard to think and speak simultaneously. The night before, he talks to himself and records what he wants to say, then uses speech-to-text to generate the script for the teleprompter. Notion does all-hands every other week and AMAs every other week, totaling weekly company-wide meetings.
Ivan wakes up at 7 a.m., makes coffee for his wife, does thinking and writing at home in the morning, then comes to the office. He comes to work every day. He's trying a new habit of going to the gym during the day to elevate his energy. He works in the office until around 7 p.m., has dinner with his wife, then works again until midnight. His most focused thinking time is in the morning or evening, though evenings he gets tired. He has one or two hour chunks between meetings for thinking. He does a lot of Slack, which his wife jokes is his social media.
Ivan wouldn't describe himself as humble. He prefers the word truth-seeking. He believes in pluralism — there are multiple flavors of truth in the world. He referenced the story of blind people touching an elephant to illustrate that each person only has one perspective of the truth.
Ivan Zhao describes his worldview as similar to blind people touching different parts of an elephant, where each person perceives only a portion of the whole—sometimes the tail seems like a snake, sometimes the trunk seems like a treat.
For founders at companies of around 10 employees hoping to scale to a thousand-person company, Ivan frames building a company as a social game that includes both positive and negative aspects. He notes that humor, status, and power are inherent human elements present in the CEO game, just as they appear in entertainment and sports.
Ivan explains that over the past three to four years he has realized the importance of maintaining equilibrium between his personal values and the company he wants to build. This alignment provides a more sustainable energy source. He acknowledges his competitive nature and need to compete to win, but emphasizes staying true to his values around building good tools, respecting craft, and considering the human side of things.
"I cannot just chase something or competition for competition's sake."
He notes that the value system a person holds determines where they sit on a spectrum. Some people compete purely for competition's sake without caring about what they are building, while others primarily want to be popular. He warns that if someone only cares about craft without building a business, they might as well be an artist. Finding his own equilibrium took time.
Ivan advises founders to lean into their strengths rather than trying to fix weaknesses. He acknowledges that certain requirements are unavoidable at scale, such as one-to-many communication, which might be handled through writing rather than speaking. Within those boundaries, founders should amplify what they are naturally good at. He argues that capabilities are becoming increasingly commoditized by machines, so an individual's unique point of view and lasting strengths become more valuable.
Ivan discusses Notion's transition into enterprise sales. The company made a mistake by avoiding traditional sales motions for roughly two years while attempting to design their own system from first principles. He notes that many smart people attempt to create something entirely new when they should respect proven approaches unless reinvention is absolutely necessary.
"I was trying to reinvent a new go-to-market motion. That's stupid."
He explains that each company should preserve its innovation points for only a few key areas rather than spreading innovation everywhere. Attempting to reinvent the wheel too many times leads to being spread too thin. Classic sales works, and Notion now makes it more efficient using internal tools.
Ivan describes initial missteps in building enterprise sales. The company first tried a hybrid PLG-flavored sales approach with system-thinker sales leaders, which did not work well. Pure PLG gives sellers the luxury of customers already wanting to buy, making it easy to fulfill demand. However, this environment makes it difficult to motivate sellers to do harder work when easier opportunities exist.
The solution came from respecting the traditional sales system while pairing complementary leadership. Erica, former CRO of GitHub, served as the system thinker, while Pvesh acted as the "meat eater"—the rah-rah, high-energy sales leader. This pairing enabled sellers to maintain high energy while respecting Erica's overall system. The approach began working effectively in the past year and a half.
Ivan reports that Notion's culture has remained receptive to different types of people, preventing any organ rejection of the sales team. The engineering team works well with sales rather than resenting compensation differences or high-energy styles.
Ivan embraces the term "cult leader" when applied to CEOs. He views a company as similar to a religion: a point of view and value system projected to the world through business and product. He cites the Catholic Church as one of the most successful companies in history, lasting 2,000 years with rituals, business models, a great founder in Jesus, and a great head of sales in Paul. The church also employed referral systems and viral marketing.
"A company is it's sort of a religion."
Ivan argues that religion, belief systems, and value systems tap into fundamental human nature—people want to believe in something to find meaning and purpose. Companies provide camaraderie similar to sports or wartime experiences. He notes that the first four words in "culture" spell "cult." Notion previously aimed to make all-hands meetings so inspirational that attendees would well up or cry, reflecting significant effort put into these gatherings.
The host reflects on Ivan's authenticity, referencing Oscar Wilde's quote "Be yourself, everyone else is taken." He observes that companies reflect their CEOs and encourages founders not to be afraid to be themselves.
A second takeaway addresses companies stuck below 500 million in revenue while growing around 30%. The host recommends a refounding moment, noting it typically involves a big step back and two steps forward, with short-term numbers appearing worse. At HubSpot, a similar refounding involved creating a new independent company inside the existing organization, similar to how Apple operated historically. He suggests founders in stuck situations should take risks, as Ivan transformed Notion from a traditional SaaS company into a fast-moving AI company.
The final takeaway introduces a potential new playbook for building and running companies, distinct from approaches used over the past 30 years. Ivan and Jack Dorsey are iterating on new models suited to the AI-native era. Historically, companies have used triangle-shaped org charts since the Roman Empire—an inefficient but previously optimal structure. The new model resembles a circle with centralized context in an AI system at the center. Initially humans make decisions, but over time more decisions transfer to the AI as it receives increasing context. This approach requires legible systems and recorded information. Companies adopting this method move faster, require fewer people, and operate with shorter planning cycles, turning one-way doors into two-way doors due to reduced execution time.
Keep Sequoia Capital in your library
Save the videos and channels worth coming back to, and find them again in one place.
