Living Free: How Notion Rebuilt Itself as an AI-Native Company

Notion’s Second Act

Notion isn’t a startup anymore. It’s a company with hundreds of employees, over $600 million in annual recurring revenue, and a valuation north of $10 billion. But founder Ivan Zhao doesn’t think that’s a reason to coast. In fact, he’s convinced that the only way to survive the AI wave is to be willing to rebuild the whole company—maybe twice.

Notion started in 2013 as a docs tool, then grew into a flexible workspace for notes, wikis, and project management. By 2021, it was a unicorn. By 2025, roughly half its revenue came from AI products. That’s not an accident. It’s the result of two deliberate “do-overs.”

The First Rebuild: Kyoto, No Distractions

Notion’s early years were rough. The team couldn’t find product-market fit, and money was running out. So Zhao and co-founder Simon Last made a drastic call: fire everyone, leave San Francisco, and move to Kyoto, Japan.

Why Kyoto? Because Tokyo apartments were too small. Kyoto offered more space for less money. By renting out their SF office and home, the company actually turned cash-flow positive for the first time.

In Kyoto, it was just the two of them—writing code, eating meals, writing more code. No team to manage, no slide decks to present. Just the core question: what is this tool for, and why does anyone need it?

That stripped-down existence shaped Notion’s DNA. It’s why the product feels like a blank canvas rather than a rigid office suite. It’s also why Zhao didn’t quit and start something new. He was obsessed with a particular kind of tool—one that amplifies human thinking. Kyoto wasn’t about changing direction; it was about cutting away everything that didn’t matter.

The Second Rebuild: GPT-4 Changed the Rules

The second rebuild came in 2023. Notion was already successful, with a large team and a mature product. The easy path would have been to keep adding features, hire more salespeople, and optimize the funnel.

Then Zhao got early access to GPT-4. He realized the software world had fundamentally shifted. GPT-3 had felt like a neat toy. GPT-4 was different—it felt like a rewrite of how knowledge work could function.

Notion had actually shipped an AI writing feature two weeks before ChatGPT launched. It generated real revenue. But Zhao wanted more. He wanted autonomous agents—AI that could understand context, search across systems, and execute tasks. That proved hard.

Starting in late 2022, Notion experimented with agents. They tried different models, fine-tuning, everything. For about a year and a half, it was painful. Zhao later admitted they were too far ahead of the technology. The product vision was right, but the models weren’t ready.

That experience taught Notion something important: building with large language models is less like constructing a bridge and more like brewing beer. You can’t command the yeast to produce a specific flavor. You run experiments, see what the model can do, and then reshape your product around that capability.

From Toolbox to AI Workspace

Notion’s original insight was that users want to build their own systems. You could combine pages, databases, tasks, and notes into whatever structure made sense for your team. Traditional enterprise software gave you fixed modules: docs are docs, spreadsheets are spreadsheets. Notion gave you blocks.

That foundation turned out to be perfect for AI. Because AI doesn’t just need a text box—it needs context. The documents, meeting notes, project updates, and customer records stored in Notion are exactly what an AI needs to be useful.

So Notion repositioned itself as an “AI workspace.” It launched Notion Agents, Custom Agents, Enterprise Search, AI Meeting Notes, Notion Mail, and Notion Calendar. These aren’t just bolt-on features. They map to three strategic goals:

• Find knowledge: Enterprise Search pulls from Notion, Slack, Google Drive, Jira, GitHub, and email. It respects permissions and gives answers with sources.
• Structure communication: AI Meeting Notes don’t just transcribe—they extract decisions, action items, and next steps, turning a meeting into living knowledge.
• Automate work: Agents can create workflows, update databases, and connect to external tools, letting AI actually do things, not just suggest them.

If the old Notion let you build a workspace, the new Notion wants AI to work inside that workspace.

Organizing Like a Jazz Band

Zhao has a favorite metaphor: Notion should be a jazz band, not a marching band. He doesn’t mean no hierarchy. He’s realistic—hierarchy is part of human nature. But a jazz band has structure while still allowing improvisation. Players respond to each other, pick up new melodies, and create something together in real time.

AI demands that kind of agility. Product roadmaps can’t be locked six months ahead. Market shifts happen weekly. Financial planning still matters—you need to know your burn rate—but product strategy must be adaptable.

That’s why Notion reorganized its product teams. Designers, engineers, and product managers now sit together early, iterating directly on model outputs and user feedback. Job titles matter less. Getting things done matters more.

Hiring for Taste and Initiative

Notion’s hiring has changed too. Zhao argues that AI has flattened many basic skills. Writing, coding, research—these are becoming commodities. What’s rare now is taste and initiative.

Taste isn’t about aesthetics. It’s knowing what “good” looks like and having an opinion about the right direction. Initiative is the willingness to move forward without waiting for permission, to experiment in the face of uncertainty.

So Notion looks for energy, curiosity, optimism, and the ability to act. They use a “barbell” approach to engineering: a few very senior architects paired with young, hungry engineers. The seniors provide direction and technical judgment; the juniors execute with AI tools. It’s a way to train the next generation without slowing down.

Even sales hiring changed. The first round isn’t a résumé review. Candidates must build something with Notion and send a link. Can you actually use the tool to solve a problem? That’s the question.

Marketing? It’s Not One Department Anymore

Notion also blew up its traditional marketing department. The old model—one CMO overseeing brand, product marketing, content, and growth—was too slow. Product changes were happening too fast for information to flow through layers.

Now marketing is split into two tracks. One is storytelling: product narrative, content, community, and social. This team sits close to the product because they need to explain something that changes every week. The other track is demand generation: leads, sales support, and growth. That team sits with revenue.

Notion’s early growth was community-driven. People shared templates on YouTube, built workflows, and spread the tool organically. But selling to enterprises is different. Customers want procurement processes, security reviews, and human salespeople. Zhao admitted Notion once tried to reinvent enterprise sales from first principles, and it failed. People still want to talk to a person when they’re spending serious money. So Notion focuses its innovation on the product, not on disrupting sales.

AI Isn’t Just About Cutting Costs

Sure, AI products come with real costs—every API call burns money. Notion thinks about that. But the bigger opportunity isn’t doing the same work with fewer people. It’s changing how knowledge work happens.

Knowledge management has always been hard. Information lives in a dozen systems. People have to search, organize, and maintain it. Most company wikis end up as static graveyards. AI can change that. If your knowledge is stored, permissioned, and connected, an AI can search, summarize, compare, and act on it.

Notion has a head start because teams already keep their projects, docs, and decisions there. The context is already in the system. AI just makes it accessible.

What Founders Should Actually Do

Zhao’s advice to other SaaS founders is blunt: stop reading reports and watching videos. Go use the models yourself. Build something small. Put AI into your product or your internal workflows. Only then will you see which paths open up.

He sees a company as an entity searching for a path in a market. A new technology like LLMs opens routes that didn’t exist before. But those routes aren’t in tech blogs or competitor press releases. They show up in real usage—your customers, your team, your own experiments.

For mature companies, the trap is inertia. The processes, departments, and metrics that made you successful can also make you slow. You keep optimizing the old product, the old sales motion, the old org chart. But if you were starting from scratch today, would you build the same company? Notion’s answer is clearly no.

So the real question for any leader is: are you willing to rebuild? Not rename, not add a feature, but rethink everything under new technological conditions. That’s what “living free” means in business—freedom from your own success.