Notion launches AI Agents, wants to leave the SaaS era behind
It is the favourite productivity app of many start-ups, and of course, in 2025, it cannot do without agents: So Notion has announced the launch of AI Agents, a new category of artificial intelligence assistants designed to execute tasks autonomously within the company’s collaborative workspace platform. The announcement was made at the company’s annual Make with Notion conference in San Francisco on September 18, 2025.
The AI Agents represent a significant evolution from traditional AI tools that primarily suggest next steps, instead functioning as integrated team members that complete end-to-end workflows. The system addresses what Notion identifies as a key productivity challenge: knowledge workers spending excessive time on coordination, documentation, and repetitive information processing. Existing Notion AI subscribers will automatically gain access to the newly released features. Later this year, the company with more than 100 million users plans to launch custom agents, which might come with additional pricing.
Key Capabilities
The AI Agents can perform complex, multi-step operations including creating and editing pages, managing databases, and generating structured content. Unlike simple chatbot interfaces, these agents develop execution plans and carry them out across multiple components of the Notion workspace.
„Teams are ready for AI that actually completes tasks,“ said Mick Hodgins, General Manager EMEA at Notion. „The MIT State of Gen AI and Business Report shows 95% of AI projects deliver no financial gain – poor AI integration costs money.“
„Notion agents are personalized AI teammates that sit directly in your workspace. The differentiator is they don’t just prompt you about tasks – you can assign tasks and they’ll do the work. They complete tasks end-to-end within your workspace, like editing multiple pages or updating databases. Give your agent a goal and watch it build and complete workflows that would take days on other platforms“, says Hodgins. „They can search your Notion workspace, connected apps like Slack and Google Drive, and the web. You can set up a profile page telling your agent how to behave, including shortcuts, response preferences, and content references.“
The agents integrate with external tools including Slack, Google Drive, Microsoft Teams, and SharePoint, while also conducting web research to gather comprehensive information for task completion. All operations respect existing user permissions and access controls.
Personalization and Learning
Users can customize their AI agents through profile pages that establish behavioral preferences, response styles, and content references. The system includes a memory function that allows agents to learn from interactions and update their profiles automatically, though users maintain manual editing control.
The platform also supports the Model Context Protocol (MCP), enabling external AI tools like Cursor and Claude to access Notion workspaces directly.
Industry Context
The launch positions Notion’s evolution from what the company terms the „SaaS era“ to the „agent era.“ This follows the platform’s progression from document-focused tools (Notion 1.0) through database integration (Notion 2.0) to autonomous AI assistance (Notion 3.0). Notion competes not only with Microsoft Office and Google Workspace, but also with Confluence, Jira and Asana. To position itself for the future, AI integration is actually a must.
Hodgins highlighted his preferred use case: „I’ve set a scheduler that reviews my projects weekly within my workspace and creates a neat to-do list showing where I need to focus. This weekly scheduler agent keeps me focused on priorities and saves huge amounts of time.“
Technical Implementation
The AI Agents utilize multiple large language models, including systems from OpenAI and Anthropic. According to Hodgins, Notion uses „a variety of LLMs… We’re keen to ensure any LLM we integrate enhances customer experience.“
The agents are immediately available to existing Notion AI subscribers through browser and desktop applications, with mobile versions planned for release soon. Companies including Parloa, Ramp, and Vercel are already implementing the technology.
Future developments include specialized agent creation capabilities, allowing organizations to build custom AI assistants for specific workflows that can operate on automated schedules or triggers without manual prompting.
The announcement reflects broader industry movement toward autonomous AI systems that can complete complex workplace tasks rather than simply providing information or suggestions.
