You are here:
Einstein - Agent Analytics Control
Activates the continuous collection and visualization of autonomous agent interactions, enabling performance tracking and ROI measurement.
Control Name
Einstein Trust Layer - Agent Analytics
Control Overview
Activates the continuous collection and visualization of autonomous agent interactions, enabling performance tracking and ROI measurement.
Description
Enables the capture of template-specific performance data. This includes how often a template is triggered and the associated feedback reasons.
Recommended Configuration
Go to Agent Analytics in Setup. Select the appropriate dashboard template. Make sure that the Tableau Next Limited Consumer and Data Cloud User permission sets are assigned to stakeholders.
Security Impact
Identifies broad patterns of agent failure or escalation spikes that could indicate a sophisticated prompt injection attack or a systemic logic error affecting thousands of users.
Business Impact
Quantifies the value of AI by measuring the Containment Rate (sessions resolved without human intervention) and Deflection Rate, directly impacting head-count planning, and budget allocation.
Security Risk If Not Configured
Executives and admins remain blind to aggregate agent behavior. Without trend analysis, potential data exfiltration or brand-damaging toxicity trends can go unnoticed for weeks.
Threat Scenarios
Unmonitored Escalation: An agent consistently fails to resolve specific security-related queries, leading to high abandonment and poor customer trust. Sentiment Decay: A stealthy change in an external LLM causes the agent to become increasingly aggressive or unhelpful.
Estimated CVSS Score Range
Critical (9.0–10.0).
Risk Impact Considerations
High impact for companies scaling AI across large teams where manual spot-checking of AI interactions is no longer feasible. With higher usage, implementing regular monitoring of different metrics a crucial step to result in the full benefit of this feature.
Higher Risk When
Agentforce is active but the metrics are not captured in a readable format, leaving data trapped in raw objects without visualization.
Low Risk When
Agentforce metrics are reviewed via custom dashboards, and automated alerts are set up to trigger when the toxicity rate or escalation rate exceeds a 5% threshold.
Business and Integration Considerations
Tracking the deprecation timeline of current capabilities, for example, Legacy Agent Analytics is scheduled for retirement in May 2026.
Security Health Review Guidance
Security Health Review verifies that the Analytics dashboard is not just "Enabled" but "Active" (receiving data) and that "API Enabled" is set to true on analyst profiles to make sure dashboard cards load correctly.
Who Is Impacted
Support directors, sales operations, AI analysts, and Salesforce admins.
