Merit
Sign in
DiscovereCommerceNotioncom
eCommerce

How Rakuten France turned every team into AI builders with Custom Agents

63/100validation score
80+
Custom Agents built
58000+
workflows automated
50%
backlog health improvement (some teams)

Rakuten France built a scalable, self-learning AI automation platform using Notion Custom Agents, enabling any employee to create their own agents in 10–20 minutes. With 80+ agents running company-wide and 58,000+ workflows automated, the four-person AI team shifted from building to enabling—achieving a 50% improvement in backlog health.

Our role isn't just to build automation solutions for the company—it's to act as a catalyst and spread best practices to every team.

Clément Caillol, CPO

If a non-technical team thinks Custom Agents aren't for them, it's actually the opposite—they're built for non-technical people.

Sidney Golstein, AI Lead

In 10–20 minutes, a new agent can be live. The simplicity of creation is the key point.

Sidney Golstein, AI Lead

I didn't have to sell Custom Agents. Others saw what the product team was getting every morning and asked if they could have it too.

Sidney Golstein, AI Lead

Custom Agents helped us move beyond individual AI experiments and into a real, scalable way of working.

Clément Caillol, CPO
Hero image

Hero image

1

About the Vendor

Notion is an all-in-one productivity and collaboration platform that combines notes, databases, wikis, and project management. Its Custom Agents feature allows users to build AI-powered automation workflows in plain language, without requiring technical expertise.

2

About the Brand

Rakuten France is the French arm of the global Rakuten e-commerce marketplace. The company operates with a small, four-person AI team and embraces the Japanese principle of "shikumika" (systematization) to drive consistent, scalable processes across all departments.

3

Challenge

Rakuten France's four-person AI team couldn't scale automation across the entire company on their own. Previous no-code tools failed to stick—when the person who built a workflow moved to another role, the automation would break. The company needed a durable, democratized approach to automation that didn't create a central bottleneck, and where any employee, technical or not, could build and maintain workflows independently.

4

Solution

Rakuten France adopted Notion Custom Agents as the foundation for company-wide automation. AI Lead Sidney Golstein built a full internal platform within Notion, including: Shared Operating Procedures (SOPs) that every agent references; the Rakuten Agent Builder (a Custom Agent that creates other Custom Agents via plain-language chat); a Friendly Onboarder agent for new users; an Agent Registry database for discoverability; Agent Logs for transparency; and the Delicious Improver agent that reviews logs and proposes SOP updates. This created a self-learning, closed-loop system.

5

Execution

Sidney hosted workshops where every team built their first agent live. He created the Notion Agents Playground as a guided path for non-technical employees. For the product team, he built Plan Snitch, Run Snitch, and Release Snitch agents that scan backlog tickets nightly, and Sprint Lebowski, which posts tailored daily reports to each team's Slack channel. Adoption spread organically as teams saw colleagues benefiting and requested the same setup. The AI team's role shifted from building agents to maintaining the platform that enables others to build.

6

Results

Rakuten France now runs 80+ Custom Agents company-wide, the majority built by employees outside the AI team. Over 58,000 workflows have been automated. Some teams, including the product team, saw backlog health issues drop by 50%. Adoption spread organically after teams witnessed the daily Sprint Lebowski Slack reports, eliminating the need to "sell" the solution internally. The AI team transitioned from builders to platform enablers, and the system is self-improving through automated log review and SOP updates.

Tech Stack

NotionNotion Custom AgentsSlack

Competitive context

Reasons for switching: Previous no-code tools broke when workflow owners changed roles, Automations were not durable or maintainable without the original builder

Validation breakdown

Audited against Merit's standard
Evidence quality67/100

How well the outcomes are documented and traceable.

Authority signals57/100

Brand recognition, client relationship, and external references.

M
Merit Validation Team✓ Verified
Independent audit · Merit standard

Data source: declared · none