Make with Notion 2025: How Notion AI Changed Work at Notion

Notion AI agents are moving work from “chatting” to “doing,” and Notion says that shift is already reshaping how teams operate internally. After rolling out a more capable personal agent and introducing custom agents, Notion reports rapid adoption—836 custom agents created in a single month, nearly one per employee—suggesting the tools are catching on even with people who previously avoided AI.

The personal agent is positioned as a one-click, Notion-aware assistant that knows the user’s workspace and can complete tasks end to end. It’s trained to work with Notion blocks, pages, and databases, then personalized through user instructions and “memories.” Most importantly, it doesn’t just answer; it plans and executes multi-step workflows—running searches, editing pages, and updating entire databases. Notion compares the experience to shaping clay: instead of manually clicking through documents and copying data around, the agent can draft an outline, push it into a Notion page, then pull supporting information from sources like Slack messages and PDFs.

Custom agents take that further by letting teams define autonomy and boundaries. Each custom agent can be configured with specific instructions, the Notion content it can access, and triggers for when it should act. The examples emphasize background execution: set a schedule (such as every Monday at 9:00 a.m.) and the agent generates a recurring deliverable without ongoing supervision. Notion also highlights scale—one agent can handle multiple tasks, and teams can run hundreds of agents together.

Several internal use cases illustrate the practical impact. “Rusty,” a marketing-focused database builder example, shows an agent pulling an agenda from a website and using it to populate a structured database with talks, speakers, times, and stages. The agent then reorganizes the database by grouping and adding views to track the day’s schedule—work that would normally take 30 minutes to an hour down to a few minutes.

“Luca” demonstrates personalization as a shortcut system. A dedicated instructions page influences how an agent behaves in new chats, including storing user-specific details like a user ID and enforcing preferred formats for explanations and research reports. A key workflow example shows how a user can paste a link in Slack and have the agent route it into a triage database, create enriched pages, and tag them appropriately.

“Ash” focuses on executive visibility through EPD (engineering, product, and design) updates. A custom agent analyzes the last seven days of project updates, extracts key risks and action items, and summarizes accomplishments and what’s shipping next—then can run automatically on a weekly trigger so the briefing arrives without manual reading.

Finally, “Smilers” is built for the environment team’s office operations. The agent responds to questions in the “office ask” Slack channel using the team’s wiki, creates and assigns tickets in a structured Notion database, and updates the wiki when new information appears. The demo shows printer-location guidance being answered from the wiki, then a new printer being added, with edit attribution recorded as agent-created changes. Notion closes with a practical starting set: use personal agents to update databases and views, analyze data, and—once custom agents are available—build Q&A bots that connect Slack and Notion knowledge so documentation stays current.