How Rakuten reached 89%
first-contact resolution
By deploying an EzyAgents AI support agent, Rakuten now handles 3× more customer requests without growing headcount, with NPS up 37 points.
Industry
E-commerce · Marketplace
Volume
50,000+ tickets/month
Use case
E-commerce customer support
Deployment
21 days
+37 pts
NPS
customer score
-55 %
Cost / contact
vs. before deployment
89 %
Resolution
at first contact
The challenge
With more than 50,000 support tickets per month, Rakuten France's customer service faced a major scaling challenge. Response times reached 48 hours during activity peaks (sales, Black Friday), first-contact resolution was stuck at 52%, and cost per ticket remained high despite optimization efforts.
Most requests concerned repetitive topics: order tracking, return policy, refunds, payment issues. Topics that are perfectly documented, yet time-consuming to handle manually at scale.
The EzyAgents solution
Rakuten integrated an EzyAgents agent directly into its ticket-processing flow, connected to its CRM, its logistics ERP and its product knowledge base. The agent operates in French and Japanese, matching the markets served.
- Automatic qualification and replies for order-tracking and return requests
- Real-time access to the customer's order data to personalize every reply
- Detection of dissatisfaction signals and priority escalation to a human advisor
- HITL approval on goodwill gestures above €50 and on refunds
- Weekly analysis of contact reasons to feed process improvement
The results
After 3 weeks of gradual rollout, the agent handled 68% of total ticket volume autonomously. First-contact resolution went from 52% to 89%. Average response time dropped from 48 hours to 4 minutes on automated requests.
Cost per contact fell by 55%, generating significant annual savings. Customer NPS rose 37 points in six months — the fastest improvement ever recorded in the history of Rakuten France's CX program.
"Our customer service handles 3× more requests without growing headcount. First-contact resolution went from 52% to 89% thanks to the EzyAgents agent."