Salesforce Launches Advanced Monitoring Suite for AI Agents
On Thursday, Salesforce announced the release of Agentforce Observability, a comprehensive set of tools integrated within its Agentforce 360 Platform that enables organizations to track and understand how their AI agents make decisions in near real time. This development targets a major challenge in enterprise AI adoption: the opacity of AI agent reasoning during live customer interactions.
AI agents are increasingly deployed to automate complex tasks such as customer support and appointment scheduling, yet businesses often lack the visibility needed to audit or optimize these systems. Salesforce’s new observability capabilities offer granular insights into each AI agent’s actions, reasoning steps, and triggered safety guardrails, aiming to build executive trust and facilitate wider AI deployment.
Addressing the Trust Deficit in AI Agent Deployment
Adam Evans, Executive Vice President and General Manager of Salesforce AI, emphasized the necessity of transparency for scaling AI: “You can’t scale what you can’t see.” The company notes a 282% surge in AI implementations among its customers, underscoring the urgency for robust monitoring frameworks capable of managing fleets of autonomous agents making real-world decisions.
Gary Lerhaupt, Vice President of Salesforce AI and head of the observability initiative, described Agentforce Observability as a “mission control system” that goes beyond simple monitoring to deliver business-specific metrics, such as engagement rates in service or lead conversion in sales, which traditional tools overlook.
Real-World Impact Demonstrated by Early Adopters
Early adopters like 1-800Accountant and Reddit illustrate the practical benefits of Salesforce’s observability tools. Ryan Teeples, CTO of 1-800Accountant, reported that Agentforce agents now operate as a 24/7 digital workforce managing tax inquiries and scheduling by integrating diverse data sources securely and autonomously.
“Observability allows us full transparency and trust with every agent interaction,” Teeples said, highlighting the importance of such capabilities during sensitive, high-volume periods like tax season. The insights provided by the system enabled 1-800Accountant to identify performance gaps and configure safety guardrails promptly, leading to over 1,000 client engagements resolved within the first day of deployment and projected capacity for 40% client growth without additional staff.
Similarly, Reddit has leveraged these tools to deflect nearly half of its advertiser support cases, with John Thompson, VP of Sales Strategy and Operations, noting that understanding the AI’s decision pathways aids in improving both issue resolution and the customer experience.
Technical Foundations: Session Tracing and Agent Fabric
Agentforce Observability is built upon two core technologies. The Session Tracing Data Model meticulously logs every AI agent interaction, from user input through reasoning and external data calls to guardrail activations, securely storing this data within Salesforce’s Data 360 platform for unified session-level visibility.
Complementing this, MuleSoft Agent Fabric addresses the complexity of managing multiple AI agents across ecosystems by providing a consolidated dashboard—the “single pane of glass”—that visualizes agent networks and their interactions comprehensively.
The tools are divided into three functional modules:
- Agent Analytics: Monitors key performance indicators and trends.
- Agent Optimization: Analyzes interaction details to detect patterns and configuration issues.
- Agent Health Monitoring: Tracks real-time health metrics and alerts on errors, with general availability scheduled for Spring 2026.
According to Pierre Matuchet, SVP of IT and digital transformation at Adecco, early testing using Agentforce Observability revealed unanticipated user behaviors and bolstered confidence in the agents’ reliability before full-scale deployment.
Competitive Edge Over Other Cloud Providers
Salesforce positions Agentforce Observability as a superior alternative to the monitoring tools offered by major cloud providers like Microsoft, Google, and Amazon Web Services. Lerhaupt emphasized that unlike these basic monitoring solutions, Salesforce’s system is included at no extra cost and provides “deeper insight than ever before” by capturing detailed telemetry and reasoning data, empowering customers to optimize agent performance effectively.
This positions enterprises at a crossroads: relying on native cloud monitoring or adopting a specialized observability platform that delivers comprehensive transparency and control over AI agents.
Scaling AI Agents: From Pilot Projects to Enterprise Production
While the rapid increase in AI deployments is promising, questions remain about the maturity of these implementations. Lerhaupt outlined a three-phase progression from initial trust-building and design to scaling, citing examples such as Williams Sonoma and Falabella, which have significantly expanded their AI workflows.
Salesforce reports over 12,000 customers across 39 countries using Agentforce, cumulatively managing 1.2 billion AI agent workflows, suggesting that production-level adoption is well underway despite the absence of detailed breakdowns between pilot and full deployments.
As companies face pressures to optimize costs without sacrificing service quality, AI agents offer potential solutions—contingent on the trust enabled by tools like Agentforce Observability.
Beyond Deployment: The Imperative of Continuous AI Monitoring
The company stresses that the AI agent lifecycle extends beyond build, test, and deploy phases. Unlike traditional software, AI agents evolve through learning and probabilistic decision-making, necessitating ongoing monitoring to detect behavioral drift and emergent failure modes.
Lerhaupt noted, “Building an agent is just the beginning. Transparency around agent behavior and outcomes is critical to optimize customer experiences and maintain trust.” Without such observability, customers like 1-800Accountant might hesitate to expand AI usage, underscoring the tool’s role in enabling confident scaling.
Trust as the Central Challenge in Enterprise AI Adoption
Interviews with customers consistently highlight trust as the primary obstacle to broader AI deployment. While AI technology and infrastructure have advanced sufficiently, executive confidence hinges on the ability to understand, supervise, and rectify AI agent actions promptly.
Salesforce frames Agentforce Observability not merely as a monitoring solution but as a management layer akin to human workforce oversight, enabling detailed supervision of AI agents’ every decision and reasoning step.
This granular visibility offers unprecedented opportunities for continuous performance improvements but also demands that organizations develop processes to translate observability data into actionable optimizations.
Ultimately, Salesforce asserts that enterprises capable of “seeing” their AI agents’ internal processes will advance more rapidly from cautious experimentation to confident, large-scale deployment, positioning observability as indispensable in the era of autonomous AI workforces.
