Notion and OpenAI bring intelligence to every workflow

Notion is a connected workspace where teams write, plan, and organize everything from meeting notes to product roadmaps. Today, it’s also a deeply AI-powered platform, used by millions to summarize content, generate writing, and ask questions in natural language across their entire workspace.

That shift didn’t come from layering AI onto existing features. It came from rethinking the core product experience and integrating OpenAI’s models throughout it.

Notion was one of the first companies to gain access to GPT‑4, and within a week, its founders had prototyped a working AI writing assistant. What started as a hackathon project became Notion AI - now core to how people get work done.

Along the way, Notion’s real-world feedback and scale has helped shape OpenAI’s roadmap, influencing improvements to routing, embeddings, and reserved capacity.

For years, Notion users described it as their “second brain”—a place to capture ideas, store knowledge, and build workflows. But until AI, it was difficult for users to act on that knowledge.

"Before, you could put your notes in Notion, but you couldn’t meaningfully act on them," says Kothari. “AI has changed that."

In 2022, Notion’s founders got early access to OpenAI’s models. Rather than participate in offsite activities, they spent a week prototyping a new version of Notion that integrated AI. 

The result was a writing tool that could instantly generate drafts, summaries, and suggestions—changing how people interacted with their content. When they demoed it internally, “a light bulb went off,” Kothari said. “Suddenly it felt like AI could meaningfully assist with knowledge work.”

Since that first prototype, Notion has evolved from a product with AI features to what Kothari describes as a “natively-AI product.” Every building block, from search to database creation to meeting notes, now leverages AI. At the heart of these experiences are OpenAI’s models, including GPT‑4o, GPT‑4o mini, and embeddings.

For example, Notion AI allows users to ask questions in natural language across their workspace and get accurate answers with citations. The feature combines OpenAI’s embeddings and models with Notion’s own retrieval systems.

“Q&A really helped us close the search gap we were seeing,” said Kothari. “Now people can just ask a question and get the correct answer, along with the source.”

To enable this performance at scale, Notion developed internal tools and infrastructure to rapidly evaluate and integrate new models from OpenAI.

“Whenever a new model comes out, our team can evaluate it within half a day,” said Sarah Sachs, Engineering Leader for AI Modeling. “We were one of the first customers to launch a user-facing product on GPT‑4o because we’d built that infrastructure.”

That real-time development cycle goes both ways. By sharing product needs and implementation insights, Notion has contributed feedback that helped shape features like reserved capacity and improvements to routing and embeddings.

For both Notion’s users and its business, the collaboration has led to meaningful results:

• Over 2 million people joined the waitlist for Notion AI within weeks of launch.
• Nearly two-thirds of users who adopt AI features become more active in the product compared to non-adopters, and three-quarters say they wouldn’t want to work without them.
• AI users save over 70 minutes a week on average through automated summarization, content generation, and task automation.
• According to a recent survey, 86% of customers would be very disappointed if Notion’s AI features were removed​.
• AI has also become a meaningful revenue stream through a paid plan in Notion’s business model.

With AI integrated deeply into its infrastructure, Notion is able to ship and iterate faster—unlocking compounding gains in both performance and capability. “When we first talked about AI features, we were building for sub-minute performance. Today, we’re operating in milliseconds,” Sachs says. “From initial alpha to general availability, we’ve seen more than a 50% improvement in latency for some features.”

Perhaps most exciting is where Notion is headed. With AI, Notion isn’t just building tools to generate or find information, but to do things with it—and eventually automate those workflows.

“AI is going from a writing assistant to a true collaborator,” says Sachs. “It’s your chief of staff. There’s no silly question, no poorly written draft, just a trusted partner that helps you operate at the top of your game.”