Make with Notion 2024: Taking Notion AI to the max (Michelle Hilzinger, Shir Yehoshua)

Notion AI’s big pitch is simple: one AI system that can search, write, and coordinate multi-step workflows across everything stored in a workspace—so people stop bouncing between separate AI tools and instead get answers and drafts in the same place. The core idea is that Notion already holds “all of life’s knowledge” in pages, databases, and documents, and the platform’s AI features use that context to deliver results that are grounded in the user’s own content.

The journey starts with two earlier building blocks. In November 2022, Notion launched AI Writer, letting users continue writing, summarize, edit, and generate content. Around the same time, general availability arrived for broader AI writing capabilities, and the company leaned on model improvements plus embeddings—numerical representations of meaning used to find related text. Those embeddings enabled Q&A: instead of surfacing pages or links, Notion AI could retrieve the “needle in a haystack,” synthesize relevant information, and answer questions using sources from within the workspace. Early feedback matched the intent: users wanted immediate answers without interrupting coworkers.

After that, the focus shifted from individual AI features to reducing tool sprawl. With many AI products appearing for chat, search, and writing, the experience of using ten separate tools became “painful,” especially when they didn’t work well together. Notion’s response was to bring context from other tools into Notion itself. Q&A was launched first, then AI connectors arrived—starting with Slack and Google Drive—so Notion AI could search and act on information beyond Notion pages.

About a month before the demo, Notion officially launched the “new Notion AI,” framed as a single AI tool that can chain multiple actions into one connected workflow. A key example: turning a day’s status check into a company-wide announcement. The system can take the latest information, draft a shareable update, and keep the workflow inside Notion rather than forcing copy-and-paste between tools.

The live walkthrough shows how that connectedness works in practice. In a hypothetical workday, a user asks what’s happening with a “Spotify AI connector” launch. Notion AI searches across sources using embeddings, cites relevant material from Google Drive and Notion, and then digs deeper into the setup flow to explain why the launch is delayed by two weeks. From there, it generates a decision document: it synthesizes the retrieved information into a structured doc with pros and cons, then extracts additional next steps from an uploaded image of handwritten notes. Finally, it drafts a short announcement to communicate the delay internally.

The second half of the demo extends the same pattern to user feedback triage. Notion AI summarizes long feature requests, categorizes them into tags automatically, and updates asynchronously as new content arrives. Once tagged, charts make trends visible so teams can prioritize what to build.

Personal use follows the same logic: Q&A can find recipes by constraints (like gluten-free), and large language model knowledge helps generate variants (such as removing cashews from a brownie recipe). The session ends with practical guidance—import more context, connect more apps, use templates and app mentions for formatting, and break tasks into smaller steps—plus a roadmap: more connectors (GitHub and Jira) and deeper support for structured data and aggregated insights beyond single documents.

Overall, the message is that Notion AI’s value comes less from one-off text generation and more from turning scattered knowledge into grounded answers, drafts, and decision-ready artifacts—then wiring those steps together through connectors and iterative workflows.