Salesforce Launches Agentforce 3 to Give Businesses Real-Time Visibility and Control Over AI Agents
Salesforce has rolled out Agentforce 3, the latest evolution of its AI agent management platform, aimed squarely at solving a key pain point for businesses: understanding what their AI agents are actually doing—and how to make them better.
Since its debut in October 2024, Agentforce has gained serious traction across industries. Companies like 1-800Accountant reported offloading 70% of chat-based administrative queries to AI during the height of tax season, while Engine shaved 15% off customer case handling times. But with adoption growing at a breakneck pace, many businesses are finding themselves flying blind after deploying AI tools.
The Command Center: A Dashboard for AI Agents
At the heart of Agentforce 3 is a new Command Center, a centralized hub that gives managers a live view of how AI agents are performing. It tracks real-time metrics like latency, error rates, and escalation volumes—making it easier to identify what’s working, what’s not, and where to focus optimization efforts.
The system uses OpenTelemetry, meaning it integrates seamlessly with popular observability tools like Datadog and Splunk—so IT teams can monitor AI performance just like they would any other mission-critical service.
CC BY 4.0
And businesses are paying attention. Slack’s upcoming Workflow Index reports that AI agent usage has surged 233% in just six months, with over 8,000 organizations now using Agentforce.
AI That Watches Itself
Beyond data collection, Agentforce 3 introduces self-optimizing intelligence. The platform doesn't just surface problems—it suggests solutions. By identifying patterns in interactions, it can recommend fine-tuned improvements to bot behaviors and responses. For busy teams drowning in conversations, this kind of autonomous feedback loop could be a game changer.
Ryan Teeples, CTO at 1-800Accountant, said the real power of Agentforce is just beginning to emerge:
“With a high level of observability, we can see what’s working, optimize in real time, and scale support with confidence.”
Solving AI Agent Connectivity With MCP
A long-standing hurdle for AI adoption has been secure, scalable integration with enterprise systems. Agentforce 3 tackles that with native support for the Model Context Protocol (MCP)—a kind of universal connector Salesforce likens to “USB-C for AI.”
Through Salesforce-owned MuleSoft and Heroku, businesses can now plug AI agents into their existing infrastructure with minimal coding and full respect for internal security policies.
Mollie Bodensteiner, SVP of Operations at Engine, emphasized the value of this plug-and-play capability:
“That level of interoperability has given us the flexibility to accelerate adoption while staying in complete control.”
Ecosystem-Driven Innovation
Perhaps the most strategic shift in Agentforce 3 is its growing partner ecosystem. Over 30 tech companies—including AWS, Google Cloud, Box, PayPal, and Stripe—have built MCP-compliant servers that extend the power of Agentforce.
These aren’t just data bridges. AWS integrations enable document analysis, audio transcription, and even key moment detection in videos. Google Cloud connections unlock access to AI tools like Imagen and Veo, plus services like Maps and BigQuery.
Healthcare is one of the sectors showing the most promise. At UChicago Medicine, VP Tyler Bauer sees AI agents as a way to streamline patient access centers by handling routine inquiries, freeing human staff for more complex or sensitive interactions.
The Road Ahead for AI Agent Management
For many companies, AI has gone from experiment to essential in under a year—but managing that complexity has lagged behind. Businesses often know how many queries the bots handle, but not whether they’re handling them well—or how to improve.
Adam Evans, EVP & GM of Salesforce AI, sees Agentforce 3 as a turning point:
“Agentforce 3 will redefine how humans and AI agents work together—driving breakthrough levels of productivity, efficiency, and business transformation.”