Salesforce Agentforce 3 Enhances AI Agent Visibility and Control

Salesforce Agentforce 3 is set to revolutionize how businesses manage their AI agents. Since its launch in October 2024, Agentforce has already made significant strides, reducing customer case handling times by 15% for Engine and handling 70% of administrative chat queries for 1-800Accountant during the tax season.

However, the latest upgrade goes beyond just performance numbers. It addresses a critical issue: the lack of visibility into AI agent activities. This is a common challenge for businesses deploying AI at a rapid pace, often without a clear understanding of what the agents are doing or how to optimize them.

Keeping Tabs on Your Agents

The centerpiece of Agentforce 3 is the Command Center, a mission control hub for AI agents. It allows managers to monitor agent performance, track health metrics in real-time, and identify areas for improvement. The system captures all agent activity using the OpenTelemetry standard, ensuring seamless integration with existing IT tools like Datadog and Splunk.

For businesses that have deployed AI tools and are unsure of the next steps, this level of visibility could be game-changing. The system not only provides data but also suggests improvements. The AI watches itself, identifying conversation patterns and recommending tweaks, which can be incredibly useful for overstretched teams.

The Connectivity Conundrum Solved

Another significant issue Salesforce is addressing is connectivity. AI agents are only as effective as the systems they can access, but connecting them securely has been a challenge for most organizations. Agentforce 3 introduces native support for the Model Context Protocol (MCP), described as “USB-C for AI.” This means AI agents can connect to any MCP-compliant server without custom coding while respecting security policies.

MuleSoft, a Salesforce acquisition, plays a crucial role in converting APIs and integrations into agent-ready assets. Heroku handles the deployment and maintenance of custom MCP servers, ensuring a robust and secure ecosystem.

Growing the Salesforce Agentforce Ecosystem

One of the most exciting aspects of this announcement is the ecosystem Salesforce is building. Over 30 partners, including AWS, Google Cloud, Box, PayPal, and Stripe, have created MCP servers that integrate with Agentforce. These integrations offer more than just data access. For example, AWS integration allows agents to analyze documents, extract information from images, transcribe audio recordings, and identify important moments in videos. Google Cloud connections tie into Maps, databases, and AI models like Veo and Imagen.

Healthcare is a particularly promising sector for AI. Tyler Bauer, VP for System Ambulatory Operations at UChicago Medicine, highlights the need for adaptable AI tools that can handle routine interactions, freeing up care teams to focus on more complex needs.

Real-World Impact

The real question is whether Agentforce 3 will help businesses manage their growing AI agent fleets. Many organizations know the percentage of queries AI is handling but struggle to identify specific shortcomings or improvement opportunities. Agentforce 3 aims to bridge this gap by providing the visibility and control needed to optimize AI performance.

Adam Evans, EVP & GM of Salesforce AI, believes that Agentforce 3 will redefine how humans and AI agents work together, driving significant improvements in productivity, efficiency, and business transformation. While the promise is lofty, addressing the visibility and control gap is a crucial step for businesses struggling to manage their AI initiatives.