Organizations have rolled out AI agents at pace, but few know what these digital workers really do or how to boost their output. Salesforce Agentforce 3 aims to close that gap by giving leaders a detailed view of agent activity.
When Agentforce first appeared in October 2024, early adopters reported significant gains across multiple industries. Engine trimmed customer case-handling times by 15 percent, and 1-800Accountant handed off 70 percent of its administrative chat questions to AI during the height of tax season.
Some users pointed to even finer gains. A national retail chain used Agentforce bots to process refund and return inquiries, cutting resolution time by more than two-thirds. Healthcare providers automated routine appointment scheduling and billing questions, freeing up staff for more sensitive tasks.
Despite these wins, many stakeholders lacked tools to explore agent behavior in depth. Agentforce 3 addresses that by offering drill-down views across workloads, user segments and individual agent profiles. Teams can set custom alerts for latency spikes, rising error rates or escalation thresholds.
The health dashboard tracks API call duration, memory usage, thread concurrency and throughput. Administrators can configure automatic scaling when demand exceeds set limits. Observability also stretches to training pipelines: developers can evaluate prompt performance, monitor drift in agent responses and run A/B tests on different conversational flows.
At the core of this release is the Command Center, a mission-control console for AI employees. It reveals performance patterns, displays live health metrics—including latency, escalation rates and error counts—and highlights areas that need refinement. A recommendation engine sits alongside this dashboard, applying analytics to spot repeated misunderstandings or dead-end queries and proposing configuration tweaks such as adjusting response templates or rerouting specific topics to specialized agents.
Every interaction is recorded through the OpenTelemetry standard, so Agentforce 3 integrates with monitoring platforms like Datadog and Splunk already in use by IT teams. This compatibility speeds deployment and leverages existing observability investments.
AI adoption continues its sharp rise. Data from the Slack Workflow Index shows agent usage up 233 percent in just six months, jumping from 15 percent of organizations in October 2024 to 50 percent by March 2025. Active agent sessions on the Agentforce network exceed 10 million interactions per week, and over 8,000 organizations have signed on during this period.
Ryan Teeples, CTO at 1-800Accountant, said: “Agentforce autonomously resolved 70% of 1-800Accountant’s administrative chat engagements during the peak of this past tax season, an incredible lift during one of our busiest periods. But that early success was just the beginning. We’ve established a strong deployment foundation and weekly are focused on launching new agentic experiences and AI automations through Agentforce’s newest capabilities. With a high level of observability, we can see what’s working, optimise in real time, and scale support with confidence.”
Agentforce 3 not only logs data but suggests refinements. The system spots conversation trends and offers configuration changes, helping teams that juggle hundreds of bot interactions hand off the review process to AI rather than tackling it manually.
Agents thrive when linked to company systems, but building secure connectors is no small feat. Salesforce added support for Model Context Protocol (MCP), dubbed “USB-C for AI.” Any server that complies with MCP can plug in without extra code, while enforcing existing security policies. MuleSoft wraps APIs into MCP-compliant services, and Heroku handles deployment, autoscaling and maintenance of custom MCP hosts.
Mollie Bodensteiner, SVP of Operations at Engine, commented: “Salesforce’s open ecosystem approach, especially through its native support for open standards like MCP, will be instrumental in helping us scale our use of AI agents with full confidence. We’ll be able to securely connect agents to the enterprise systems we rely on without custom code or compromising governance. That level of interoperability has given us the flexibility to accelerate adoption while staying in complete control of how agents operate within our environment.”
More than 30 partners have built MCP servers compatible with Agentforce, including AWS, Google Cloud, Box, PayPal and Stripe. AWS integration lets agents analyze documents, extract data from images, transcribe audio and flag key moments in videos. Google Cloud connections tap into Maps, databases and AI models like Veo and Imagen. Other integrations let agents fetch files from Box, trigger billing flows in Stripe or PayPal and update records in enterprise apps—all within a single conversation.
Healthcare emerges as a prime use case. Tyler Bauer, VP for System Ambulatory Operations at UChicago Medicine, explains: “AI tools in healthcare must be adaptable to the complex and highly individualised needs of both patients and care teams. We need to support that goal by automating routine interactions in our patient access center that involve common questions and requests, which would free up the team’s time to focus on sensitive, more involved, or complex needs.”
Visibility into AI agent performance has been a blind spot for many organizations. Most know roughly what share of inquiries their bots handle, but struggle to uncover specific issues or spot opportunities for improvement.
Adam Evans, EVP & GM of Salesforce AI, says: “Agentforce 3 will redefine how humans and AI agents work together—driving breakthrough levels of productivity, efficiency, and business transformation.”