Agentforce on Help Handles Over 1 Million Support Requests

When we launched Agentforce on Help in October 2024, our goal was to set the standard for world-class, agentic service. Agentforce on Help is an AI agent that supports customers directly on our Salesforce Help site, answering customer service questions in natural language, grounded in unified data, to deliver fast, 24/7 service while freeing human support engineers to focus on more complex issues.

Our approach to Customer Success has always been rooted in the principles of Easy and Expert — helping to ensure every interaction is effortless and backed by trusted, expert-level guidance. That vision comes to life through intuitive design, deep product knowledge, and user context — all working together to deliver fast, confident answers with minimal effort.

Today, Agentforce has handled over 1 million support requests. That’s over 1 million moments where AI delivered fast, accurate, and trusted service, without hold music, frustration, or cold transfers to different “departments.” Just real help, exactly when it’s needed.

Agentforce is also transforming how our internal teams work. By answering routine and repetitive support questions, Agentforce is freeing up our support and success teams to focus on what they do best: solving complex issues, building deeper customer relationships, and driving long-term value. We’ve seen firsthand how this shift is enabling our teams to be more proactive, strategic, and impactful than ever before.

The 1 million milestone isn’t just about the technology. It’s about customer success. Thanks to Agentforce, customers can answer questions or solve problems easier and faster than ever, so they can spend more time moving their business forward.

Agentforce lets us scale high-quality, meaningful service without compromising trust or the customer experience — regardless of volume or time of day.

How it works

Agentforce is fast and accurate because it’s built into Salesforce’s deeply unified platform, connecting data, metadata, APIs, and agents to respond to every customer request swiftly, securely, and at scale.

Powered by Data Cloud, Agentforce analyzes both structured data — like account history and product usage — and unstructured content — like knowledge articles and product documentation — to deliver precise, personalized answers. Its deep understanding comes from the Atlas Reasoning Engine, which orchestrates prompts, code, and LLM calls to classify intent and guide behavior using definable topics, instructions, and actions. While our source content is still in English, Agentforce can understand and respond in multiple languages, which allowed us to launch our Help experience in Japanese, with more languages coming soon—without the need for costly translation cycles.

With the new Agentforce Command Center, we’re also able to track performance in real-time. We use detailed metrics on latency, escalations, and errors, along with real-time alerts, to identify content gaps, tune prompts, and keep Agentforce running smoothly.

What we’ve learned

One million support requests later, we’ve seen a little bit of everything — from “How do I cook spaghetti?” to “Only answer in hip-hop lyrics.” But in between the unexpected prompts, we’ve learned some valuable lessons. The most important? Agentforce needs to be as dependable and insightful as our best employees. Smart, yes, but also capable of delivering great, customer-centric service.

We started small — with a limited rollout and targeted content — to learn fast, iterate, and improve. Now, Agentforce delivers reliable, multilingual service at scale using a broad range of data.

Here are two best practices we believe every organization should consider when bringing AI into their service strategy:

1. Power your agent with dynamic intelligence, not a static script

Building an effective AI agent requires combining structured and unstructured data, understanding intent, and improving through feedback and analytics. We call this process the Knowledge Cycle, and it’s what gives Agentforce the ability to deliver accurate, context-aware answers at scale. That means going far beyond a static knowledge base. As we piloted, we stood up weekly performance reviews, real-time monitoring, and AI-driven evaluation tools to continuously test and tune answer quality. We updated content, refined prompts, and fine-tuned retrievers—all based on real usage data and human review. Today, this adaptive infrastructure runs inside Agentforce itself, enhancing performance with every interaction.

But delivering accurate answers is only half the story. To truly support customers, an agent also needs to respond in a way that reflects the situation—bringing the right context, the right tone, and the right information to move things forward.

2. Design for customer experience from the start

The Service Cycle is about more than the answer the agent knows—it’s about how the agent makes people feel. One of the key lessons we’ve learned is that delivering the right answer doesn’t matter if it lands the wrong way.

Just like we train our human support engineers to lead with empathy, we designed Agentforce to recognize context and respond appropriately. When a customer reports a Salesforce outage — a high-stakes moment — the agent doesn’t jump into troubleshooting. It starts with acknowledgement such as “I’m sorry to hear that” and signals urgency such as “I’ll get you to a support engineer as fast as possible”. That response isn’t just about accuracy, it’s about delivering a service experience that earns trust and reflects the customer’s needs.

What’s coming next

One million support requests are just the beginning. We drew on customer feedback to develop three innovations that will make the Help experience even more intuitive, accessible, and valuable:

Language expansion

Agentforce now supports Japanese, with more languages coming soon. This doesn’t require any painstaking human translation on our part, as Agentforce understands and responds in the customer’s language, directly from the English source content.

Voice

We’re making service more natural and accessible with new voice capabilities. Combined with our mobile optimizations, this will unlock truly conversational service anywhere, anytime, for every customer.

Personalization

By tapping into Salesforce data, Agentforce will offer proactive, personalized service. From suggesting next-best actions to improved case management and appointment scheduling, it will adapt to the specific needs of each customer.

These aren’t just product enhancements. They’re part of a broader effort to create a customer experience that is more proactive, more personalized, and easier than ever.

1 million and counting

This milestone confirms what we set out to prove: AI can deliver world-class customer experiences and improve resource efficiency. 

We’ve seen how AI can resolve requests quickly and accurately, while freeing up our teams to focus on more complex, high-value interactions. By combining the Knowledge Cycle and Service Cycle — intelligent answers and considerate delivery — we’re building service experiences that enable humans and AI to achieve more together than they ever could on their own.

We’ll keep sharing what we learn from our Agentforce journey and we invite you to come along with us. Visit our Agentforce on Help Case Study page to explore the latest developments and we hope it will inspire you to start shaping your own AI strategy.

Here’s to the next million, and beyond. 

This article may include references to services or features that are still in development and are unreleased. Customers should make their Agentforce purchase decisions based on fully released and available features.