How LY Corporation is building an Agentic Enterprise that delivers top-tier service to 107 million customers

Before agentic capabilities like Agentforce, this was very difficult for us.

We serve a large and diverse customer base, but as I said, we have a limited number of employees. That meant we could only provide highly personalized service to top-tier customers. Millions of other users — people we care deeply about — received more generic service or bulk communication.

Now, Agentforce allows us to expand that top-tier experience to everyone. Our goal is to take the service we provide to top customers, allow Agentforce to learn from those interactions, and extend that same experience to customers who previously didn’t have access to it.

One example is our shopping mall platform called Yahoo! Shopping. We have sales representatives who support merchants on the platform, but limited sales resources meant we couldn’t provide hands-on support to every store owner. In the future, we believe Agentforce will help us expand that support to any merchant who needs it – regardless of their size, or store location.

Even at our size, our engineering resources are not infinite. We want our internal teams focusing on business logic that exists only within LY Corporation — not custom-building AI tools that already exist. The knowledge within our company is where we have a true competitive advantage.

That makes it more cost-effective to use a prebuilt solution from a technology partner we already trust. We’ve been using Salesforce for many years, so it was naturally the first platform we considered. And because Salesforce is already part of our environment, adopting Agentforce within that ecosystem allowed us to start quickly and build on a foundation we already trust.

Across the company, there was already a strong belief in the importance of AI. Many teams also had firsthand experience with the Salesforce Platform. Our entire Customer 360 runs on Agentforce Service, and we use MuleSoft to integrate our home-built technology with Salesforce CRM, so there was already confidence in the technology.

Another important factor was that the vision and direction Salesforce has for agentic capabilities inside CRM aligns with our corporate strategy. When your technology partner’s roadmap matches what you want to achieve, you can move much faster. Leadership understood that if we want to move quickly with AI, we need a platform that already connects our data, applications, and customer operations.

When we prioritize agentic use cases, we look at how much work AI can perform compared to an employee, relative to the time and budget required. A simple example is customer support response time. We operate more than 100 services, with over 120 knowledge and policy documents to describe and cover them all. During peak periods, inquiry volumes spike, and our service employees are strained — searching across all those systems and documents takes time, and they can’t keep up with the volume.

With Agentforce, response times could become significantly shorter. Speed is a key part of quality in the customer experience — when customers receive a fast, accurate response, they’re often satisfied in a single interaction.

But this isn’t about replacing people. Our message internally was clear: AI is not taking the jobs you do today — it’s handling the things you didn’t have time to get to. That helped remove anxiety on the front lines.

Our data foundation for customer support already existed in Salesforce – and was integrated into our internal systems – so the context needed to power AI-driven service was readily available. Customer support is one of the most standardized processes in our company, which makes it an ideal starting point for AI.

If a process isn’t standardized, even if AI replaces certain tasks, it’s difficult to understand what employees should do with the time they gain back. For example, if AI assists engineers, the impact might vary widely — better code quality for one person, faster releases for another, or improved documentation for someone else.

Customer support is different. The procedures and quality standards are clearly defined, which makes the impact of Agentforce much easier to measure: Human workload decreases and response time improves.

Honestly, the fact that from launch, we didn’t need a human-in-the loop for many FAQ scenarios — which used to drive a high volume of inquiries.

To build our FAQ agent, we connected our rich data knowledge base already in Salesforce, allowing Agentforce to reference product information, policy FAQs, and our brand voice when responding to customers via chat. Now, when a user sends an inquiry, Agentforce reasons through the request and uses our trusted context to generate a response. The Agentforce outputs are so trustworthy, we’re projecting 80% autonomous, first-touch resolution.

One interaction really stayed with me. A customer contacted us to delete the account of a family member who had passed away. Instead of returning a static instruction about account management, Agentforce first offered a message of condolence. We hadn’t written a prompt telling the AI to do that. It responded empathetically on its own. Even now, remembering that interaction makes me emotional — it showed our support team that this technology isn’t just automation, but something we can truly work together with.

Slack and the agentic capabilities infused right in. Slack is really the center of how work happens at LY Corporation. All employees use it as their primary communication tool, enabling effortless collaboration across global teams. Personally, I’m using Slack throughout the day. If I’m not in virtual meetings, I’m on Slack.

We’re already using many Slack AI features, and adoption is very high. For example, Thread Summaries and Channel Summaries are used by almost everyone, and they save a lot of time when catching up on conversations.

We also measure the impact carefully. We calculate how much work Slack AI saves per action and aggregate those results to visualize the difference between work done by AI versus a person.

Slack has been part of our infrastructure for so long that we don’t even debate the ROI internally anymore. The company simply couldn’t run without it.