Notion's Historical Reinvention: From SaaS to AI Workspace

The Quiet Revolution at Notion

Notion didn't start as the productivity giant it is today. In 2013, it was just another ambitious idea in a crowded market. But a decade later, it's a $10 billion company with over $600 million in annual recurring revenue—and half of that now comes from AI products. This isn't just a success story; it's a case study in how a mature SaaS company can pivot before it becomes a relic.

Ivan Zhao, Notion's founder, doesn't mince words. He believes the only way to survive the AI wave is to be willing to rebuild your company from scratch. And he's done it twice.

The First Rebuild: Kyoto, 2016

Notion's early years were brutal. The team had a vision but no product-market fit. Money was running out, and the despair was palpable. Zhao and co-founder Simon Last made a radical call: they laid off the team and moved to Kyoto, Japan.

Kyoto wasn't a romantic choice. It was practical—bigger, cheaper spaces than Tokyo. They rented out their San Francisco office and home, which actually made the company cash-flow positive for the first time. In Kyoto, the duo lived a monk-like routine: code, eat, sleep, repeat. With no team to manage, they had to confront the fundamental question: what problem is this tool really solving?

Zhao didn't start a new company because he wasn't chasing entrepreneurship for its own sake. He was obsessed with the idea of a tool that enhances human thinking. So the Kyoto reset wasn't about changing direction; it was about stripping away the non-essential and returning to that core vision. That's why Notion became the flexible, block-based workspace we know today—not a rigid project management tool or a simple doc editor, but a space where you can build your own system.

The Second Rebuild: The GPT-4 Moment

By 2023, Notion was a different beast. Hundreds of employees, a mature product, and a valuation north of $10 billion. The easy path would have been to keep adding features, hire more salespeople, and optimize conversion rates. But then GPT-4 changed everything.

Zhao got early access to GPT-4 and immediately saw that this wasn't just an incremental improvement. It was a fundamental shift in how knowledge work could be done. Notion had already shipped an AI writing feature two weeks before ChatGPT launched, but Zhao wanted more. He envisioned AI agents that could understand context, retrieve information, and execute tasks. But that ambition nearly broke the team.

From late 2022, they experimented with different models, fine-tuning, and customizations, but nothing worked reliably. Zhao later admitted they were 'living too close to the future.' They were a year and a half into a painful slog. This is the key difference between traditional software and AI products: building software is like building a bridge—you have a blueprint, you execute. Building AI is like brewing wine—you can't command the yeast; you have to experiment and adapt.

That realization transformed Notion's development process. The old linear workflow—product manager takes customer requests, hands to design, then to engineering—became obsolete. Now, design, product, and engineering sit together early, iterating around model capabilities and user feedback. Titles matter less; what matters is whether you can ship something that works.

From Toolbox to AI Workspace

Notion's original strength was letting users build their own systems. Unlike traditional office suites with fixed functions, Notion was a set of building blocks. AI made that foundation even more valuable because AI needs context—your documents, tasks, meetings, and decisions are all context it can use.

So Notion pivoted from being a note-taking app to an 'AI workspace.' The product now includes Notion Agent, Custom Agents, Enterprise Search, AI Meeting Notes, Notion Mail, and Calendar. These features serve three purposes:

• Find knowledge: Enterprise Search unifies information scattered across Notion, Slack, Google Drive, Jira, GitHub, and email. AI searches within permissions and gives answers with sources.
• Structure communication: AI Meeting Notes captures decisions, action items, and follow-ups, turning meetings into part of the knowledge system rather than just recordings.
• Execute repetitive work: Notion Agents let users describe a task in natural language, and AI creates workflows, updates databases, generates reports, and connects to external tools.

If old Notion was about building a system, new Notion is about letting AI work within that system.

Organizational Jazz: Not a Marching Band

Zhao describes the ideal Notion organization as a jazz band, not a marching band. That's not an anti-hierarchy manifesto—he acknowledges that hierarchy is human nature. But in a jazz band, there's structure, yet members respond to each other in real time. That's what AI-era software companies need.

The market changes weekly, not quarterly. Product roadmaps can't be set in stone for six months. Finance planning still matters, but product strategy must be fluid. The traditional SaaS playbook—separate departments for product, design, engineering, marketing, sales—works in stable times but breaks when AI blurs boundaries and changes how knowledge work flows.

Zhao uses a metaphor: AI is like steel in construction. Before steel, you could only build five-story buildings. With steel, you can build skyscrapers. AI doesn't just improve individual efficiency; it can restructure how information moves and decisions are made.

Hiring for Taste and Initiative, Not Just Experience

AI has leveled the playing field for many basic skills. Writing and coding are now accessible to more people. So Notion's hiring criteria shifted. They now value two things above all: taste and initiative.

Taste isn't about aesthetic snobbery; it's knowing what good looks like and why. Initiative is the willingness to act, even in uncertainty. Notion hires fewer mid-level specialists and more of a barbell structure: a few very senior architects paired with young, energetic engineers. The seniors provide direction and technical judgment, while the juniors execute with AI tools. This trains the next generation and scales senior talent.

Designers and product managers must now be hybrids. Designers need to code; PMs need to prototype and test models. Even sales hiring changed—the first interview is to build something using Notion and share a link. It's about showing you can solve problems with tools, not just talk about past experience.

Marketing Deconstructed

Notion dismantled its traditional marketing department. The reasoning: in a fast-moving product environment, a centralized marketing org creates too much lag. They split it into two functions: storytelling (product narrative, content, community) that sits close to the product, and demand generation (leads, growth, sales support) that sits close to revenue.

Notion's community-driven growth—users sharing templates on YouTube and social media—was great for consumer adoption, but enterprise sales require a human touch. Zhao admits he once tried to reinvent sales from first principles, but realized that enterprise buyers want to talk to people, not robots. Innovation should be focused where it matters—product and AI workflows—not in every department.

The Real Impact: Beyond Cost Cutting

AI in Notion isn't just about doing more with less. It's about changing the fundamental nature of knowledge work. Company knowledge has always been fragmented across systems, and manual curation often fails. With AI, if knowledge is recorded and permissioned, AI can search, summarize, and act on it. The knowledge base becomes something AI reads, not just humans.

Notion's decade of accumulated user context—projects, tasks, meetings, documents, comments—becomes a moat. Teams don't need to upload data for AI; it's already there. AI just makes it usable.

Lessons for Founders and Teams

Zhao's advice to other SaaS founders: get your hands dirty. Don't just read reports or watch videos. Build with AI yourself. The paths AI opens are unique to your product, customers, and team. You can't find them in a competitor's keynote.

For established companies, the risk is being trapped by past success. The processes, departments, and metrics that got you here can hold you back. The question is: if you started from scratch today, would you build it the same way? Notion's answer is clearly no.

AI transformation isn't just adding a chatbot. It's rethinking how users accomplish tasks, how your product is structured, and how your team collaborates. It means embracing uncertainty and iterating fast. It means hiring for taste and initiative, not just credentials. And it means founders must lead the charge, not delegate it.

The good news? Old companies can beat new ones. But only if they're willing to rebuild themselves—again and again.