Cover of Yahoo Japan browser on mobile next to headline claiming 55% of emails resolved agentically

AI agents reply to LY Corporation’s service emails, reducing rep burnout

Agentforce instantly resolves 48,000 routine email inquiries per month, filters 99.9% of spam, and routes complex cases with full context.

The Results

55 %

of emails resolved agentically

48K

email inquiries resolved/month by Agentforce

99.9 %

spam emails caught agentically

700 reps, millions of requests

LY Corporation powers daily digital life for everyone in Japan and throughout the region, connecting people through messaging, search, ecommerce, and digital payments. LY Corporation’s contact center comprises seven service locations with 700 front-line customer service reps, together handling 3.3 million inquiries each year via email, phone, and chat.

Many requests were routine — like resetting a password or asking about payment refunds — yet the workload still piled up. LY Corporation receives 130,000 customer emails, most handled manually, plus 40,000 spam emails every month — enough to require 10 months of work for one employee to review and close.

Adding to the strain, reps regularly dealt with hostile messages like, "Fix this now!" Hours spent managing harassment and escalations took a toll on morale, mental health, and performance, creating an opportunity to expand self-service and help reps focus on complex customer needs.

Autonomous help for busy reps

To ease the pressure on their contact center, LY Corporation introduced agentic replies in Agentforce Service designed to autonomously handle the most common email inquiries and filter out the bad actors.

When a customer submits a web form or email, Agentforce analyzes the message, classifies the request, and generates a tailored reply using LY Corporation’s knowledge and past case patterns. Then it sends the reply, resolving the request without a rep ever needing to step in.

Built-in toxicity detection shields reps from contentious conversations by identifying hostile or abusive language and preventing those interactions from reaching the service team. All messages are still captured for sentiment analysis, and customers receive a polite response explaining that live assistance isn’t immediately available while offering alternative channels for support.

When a case requires human judgment, Agentforce routes it to a service rep with the full interaction history, intent classification, and recommended next steps.

“When we introduced AI email replies to customer support, the key was that humans weren't able to do everything they wanted to do anyway. We could only cover a part of it,” said Norihiro Hattori, Executive Officer and CIO. “Our message was never ‘AI is taking the jobs you do now,’ but rather, ‘AI will handle the things you couldn’t get to, so you can focus on higher-level customer experiences."

Accuracy powered by unified data

Each time an email comes in, Agentforce initiates a workflow behind the scenes in Service Cloud:

1. Classify the request. Agentforce analyzes the message to determine the customer’s intent — for example, whether the customer is asking about refund procedures, login issues, or an auction transaction.

2. Ground the response in LY Corporation’s trusted context. Agentforce pulls from over 500 knowledge articles indexed in Agentforce Service, along with real-time customer data from CSHUB, LY Corporation’s internal data hub connected via a secure MuleSoft API. This includes details such as subscription status and transaction history, ensuring the reply reflects the customer’s actual account situation.

3. Generate and send the reply. The workflow is built around a custom object called CS Response Flow that provides the context for the prompt template Agentforce uses to generate a tailored reply to the customer.

4. Escalate when human judgment is needed. If the inquiry falls outside the workflow’s capabilities, the system routes the message to a qualified rep. The assigned rep receives the original message, intent classification, and related customer data, allowing them to step in with full context and respond quickly.

55% of emails resolved by agents

Agentforce email replies have cut support rep workloads nearly in half and absorbed routine requests so reps can focus where they’re needed most. Beyond prioritizing their time, Agentforce has improved day-to-day work. By filtering abusive messages and handling nonproductive exchanges, such as prolonged back-and-forth with customers unlikely to reach resolution, Agentforce eliminates interactions that previously took up to 30 minutes of rep time. The result is a safer and less stressful work environment where reps spend more time solving real customer problems.

The customer experience has also improved, with email response times dropping from up to 72 hours to near-instant. Agentforce now processes roughly 2,900 inbound emails each day with a 55% resolution rate. That translates to about 48,000 emails resolved agentically each month, helping teams move through requests faster while improving customer satisfaction.

With routine emails handled automatically, service teams now have more capacity for high-value work, from improving internal processes to completing advanced training and contributing to new initiatives. LY Corporation’s long-term goal is not simply to reduce rep workloads, but to redirect skilled employees into new roles, services, and departments as the business continues to grow.

Customer Story

Agentic service to resolve 80% of customer FAQ for LY Corporation.

Agentforce resolves 25,000 customer FAQs a month on the Yahoo! JAPAN portal, with every answer grounded in trusted knowledge, order data, and case history.

Learn more

Why LY Corporation chose Salesforce

Structured AI development

Using Prompt Builder, an LY Corporation Salesforce admin built and deployed a spam detection workflow with 99.9% filtering accuracy. The workflow combines structured and unstructured data across systems to power automation without heavy development.

A system of context with MuleSoft

Before MuleSoft, service reps had to switch between Agentforce Service and LY Corporation’s internal data hub to find customer details. MuleSoft pulls the right information automatically into Agentforce Service through secure APIs — with built-in authentication, access controls, and encryption — so reps can see a full customer profile and engagement history from one workspace.