How Rakuten Uses Claude Managed Agents to Cut Release Cycles from Quarterly to Biweekly

Rakuten’s scale is unusual even among technology giants: more than 70 distinct businesses spanning e-commerce, travel, fintech, and digital content, all operating under one corporate umbrella. When the company committed to an “AI-nization” strategy—making AI central to every business function—it needed infrastructure that could match that ambition. Deploying a chatbot or a code assistant was not enough; Rakuten wanted autonomous agents capable of running complex, multi-hour workflows with persistent memory and cross-system access.

Before adopting Claude Managed Agents, Rakuten’s AI use was largely confined to software development. Teams worked with short-context interactions, and any learning from one session was lost in the next. Scaling AI beyond engineering meant rebuilding workflows from scratch for each new team, which consumed time and engineering capacity the company did not want to spend on infrastructure.

Rakuten implemented Claude Managed Agents as a shared platform for engineering, product management, sales, marketing, and finance. Agents connect to Slack and Microsoft Teams for task intake, operate with multi-hour execution windows suited to complex workflows, and maintain persistent memory across sessions—so a lesson learned in one run propagates instantly to future runs. The engineering lead Tanapat Ratana built a self-improving production exception analysis agent that non-technical teams can now run independently.

The numbers tell the story clearly. Major software releases moved from once per quarter to every two weeks. Critical errors dropped 97% in the initial rollout. Feature delivery time fell from 24 days to 5. Cost and latency improved by more than 30% without any loss of output quality. Product manager Shoko Sakamoto now builds FinOps dashboards across cloud platforms and manages production monitoring cycles that previously required dedicated engineering sprints.

Rakuten’s leadership frames this not as automation but as a fundamental shift in how work is assigned. Employees delegate goals, not tasks. The agents reason, adapt, and self-correct. Yusuke Kaji, General Manager of AI for Business, sees this as the beginning of a broader pattern: “Technical innovation starts with a small group of users and then quickly scales to change the world.” For Rakuten, that world is already changing.