Notion's Custom Agents Help Rakuten France Build 80 Agents With a Four-Person Team.

Rakuten France's four-person AI team found a way to ship 80 agents without building 80 agents. They built one Custom Agent that lets everyone else create their own.

The result: 80+ agents and 58,000+ automated workflows, all stemming from a single reusable template. Notion announced the case study on LinkedIn, highlighting how Custom Agents can scale productivity across an organization without requiring a large AI team.

Rakuten France didn't build agents one by one. Instead, the AI team created a Custom Agent designed to help other employees build their own agents. This meta-agent approach turned agent creation into a self-service tool, enabling teams across the company to automate workflows without waiting for engineering resources.

Custom Agents in Notion work by combining tools, databases, and instructions into reusable templates. Once a template is built, anyone can duplicate and customize it for their specific use case. As previously reported, Notion added duplication functionality to make this scaling pattern easier.

The approach relies on Notion's Developer Platform, which launched earlier this year with tools for syncing external data, building custom integrations, and extending agent capabilities. Developers can write Workers that continuously sync data from tools like Zendesk or Linear into Notion databases, then build Custom Agents that act on that data.

The numbers tell the story. A four-person team generated 80+ agents and 58,000+ automated workflows. That works out to roughly 14,500 workflows per person, a level of leverage that traditional software development rarely achieves.

This mirrors a pattern emerging across AI-powered platforms: small teams using meta-tools to enable large organizations. Instead of hiring dozens of automation specialists, Rakuten France invested in a template that turned employees into agent builders.

The Custom Agent template likely includes pre-built tools, database schemas, and instructions that guide users through common automation patterns. Users can customize variables like data sources, notification rules, or approval workflows without touching code. Notion's schema builder and tool framework make this possible by abstracting technical complexity into declarative configurations.

The case study demonstrates that Custom Agents are production-ready for enterprise use. Rakuten France is a subsidiary of Rakuten Group, a Japanese conglomerate with over 30,000 employees, so the scale test is significant.

For teams considering Custom Agents, the takeaway is clear: invest in templates, not individual agents. Build one good meta-agent or workflow framework, then let users customize it. This is especially relevant for operations teams, customer support, and sales enablement where workflows follow similar patterns across departments.

Notion has been expanding Custom Agents rapidly, adding MCP connectors and more powerful models like Claude Fable 5 for multi-step workflows. Combined with the Developer Platform's sync capabilities, teams can now build agents that pull live data from external tools, generate reports, update databases, and trigger notifications, all without leaving Notion.

The Rakuten France example shows what's possible when you combine those capabilities with a template-first approach. Anyone can create Custom Agents, which means anyone can scale their productivity without waiting for engineering resources.