Back office ops · Production

Rakuten deploys Claude Managed Agents across business functions, cutting critical errors by 97% in pilot

The problem

Before Managed Agents, Rakuten's engineers had to build their own agent infrastructure for persistent compute, memory, and storage, investing significant effort in scalability and reliability work that was not their core differentiator.

Workflow diagram · grounded in source
1
Task assigned via collaboration tool
trigger
“we integrate agents with Slack, Microsoft Teams, and our own Kanban-style task system, where users create and assign tasks to agents”
2
Agent collects user feedback and creates tickets
ai_action
“we collect user feedback through an agent. The agent chats with users to understand their needs and pain points, then creates tickets.”
3
Ticket triage
human_review
“Another agent or our human colleagues triage the tickets.”
4
PRD and prototype development
ai_action
“we work with agents to finalize the PRD, wireframe, or prototype. We run the iteration quickly until we meet our success criteria.”
5
Production exception investigation
ai_action
“Tanapat designed an agent that investigates production exceptions, delivers root cause analysis to Slack, and self-improves from feedback.”
6
Agent memory self-improvement
feedback_loop
“Our agents with memory remember what went wrong in past sessions and avoid repeating those mistakes.”
Reported outcome

With Claude Managed Agents, Rakuten deploys specialist agents within a week across engineering, product, sales, marketing, and finance.
In a pilot, initial critical errors dropped by 97% and cost and latency fell by more than 30% without any loss in output quality. One product manager now oversees major releases every two weeks, compared with a full quarter previously.

Reported metrics
Initial critical errors reduction (pilot)97%
Cost and latency reduction (pilot)more than 30%
Output quality change (pilot)without any loss in output quality
Specialist agent deployment timewithin a week
Show all 5 reported metrics
initial critical errors reduction (pilot)97%
cost and latency reduction (pilot)more than 30%
output quality change (pilot)without any loss in output quality
specialist agent deployment timewithin a week
release cadence improvementmajor releases every two weeks when it used to take a quarter
Reported stack
Claude CodeSlackMicrosoft Teams
Source
https://www.anthropic.com/customers/rakuten-qa
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

With Claude Managed Agents, Rakuten deploys specialist agents within a week across engineering, product, sales, marketing, and finance.

What tools did this team use?

Claude Code, Slack, Microsoft Teams.

What results were reported?

Initial critical errors reduction (pilot): 97%; Cost and latency reduction (pilot): more than 30%; Output quality change (pilot): without any loss in output quality; Specialist agent deployment time: within a week (source-reported, not independently verified).

How is this back office ops AI workflow structured?

Task assigned via collaboration tool → Agent collects user feedback and creates tickets → Ticket triage → PRD and prototype development → Production exception investigation → Agent memory self-improvement.

Related back office ops cases