Loading
Upcoming Mandatory Changes to Public Key Infrastructure (PKI)Read More
Salesforce Enforces New Security Requirements in Summer 2026Read More
Agentforce and Einstein Generative AI
Table of Contents
Filter by (0)Add
Select Filters

          No results
          No results
          Here are some search tips

          Check the spelling of your keywords.
          Use more general search terms.
          Select fewer filters to broaden your search.

            Search all of Salesforce Help
            Search all of Salesforce Help
            Data Model for Agent Optimization

            You are here:

            1. Salesforce Help
            2. Docs
            3. Quickstart Your Einstein Generative AI Solution

            Data Model for Agent Optimization

            Agent Optimization extends Session Tracing Data Model (STDM) capabilities by provisioning additional data model elements that track session moments and user intent. An LLM identifies, clusters, and tags these moments. The resulting information enables targeted queries and insights into user engagement metrics, moment duration, and response relevance quality scores.

            Required Editions

            Available in: Enterprise, Performance, and Unlimited Editions with an Einstein for Sales, Einstein for Platform, Einstein for Service, Einstein 1 Service, or Einstein GPT Service add-on. To purchase add-ons, contact your Salesforce account executive.

            Agent Optimization Data Model Elements

            When you turn on Agent Optimization, the system provisions a data model in Data 360 to track agent interactions and moments. Elements include a data stream and its target Data Lake Object (DLO), and a semantic data model (SDM) with:

            • Entities such as DMOs and Logical Views
            • Relationships that define the joins and connections among different SDM entities.
            • Dimensions and Measurements

            This illustration shows key elements.

            An entity relationship diagram (ERD) showing data model elements for Agent Optimization.
            DLO/DMO Description
            AiAgentMoment Represents a group of consecutive interactions that share the same intent, collectively forming a coherent moment within the user experience.
            AiAgentMomentInteraction A junction object indicating which interactions are part of a given moment.
            AiAgentTagDefinition Defines a tag type that you can use to annotate AI agent moments or sessions, including metadata such as the tag's name, type, or category.
            AiAgentTagDefinitionAssociation Associates a tag definition with a specific AI agent, indicating which tags are available for use by that agent.
            AiAgentTag Represents an instance of a tag definition with a specific value applied to a particular moment or session.
            AiAgentTagAssociation Associates a tag with a specific moment or session, enabling structured annotation for analysis or filtering.

            Agent Optimization Calculated Fields

            Investigate agent performance and calculate session metrics. Session tracing makes it easier to get granular insights when you join several calculated fields together. For example, you can check your agent performance in specific session moments by combining the "Unique Moments" and "Average Quality Score" fields. The "Average Quality Score" field refers to how relevant an agent's response was to a user's request.

            Dimensions are categorical or descriptive attributes that you can use to slice, filter, and group your data. For example, segmenting your agent performance data by time periods, status types, or quality categories.

            Dimensions
            Field Name Description
            Abandonment_Status_clc Indicates whether a user has left the session.
            Agent_Interaction_Latency_clc The time it takes the agent to respond to the user's request, in milliseconds.
            Deflection_Status_clc The deflection status, indicating if the session ended at the user's request.
            Engagement_Status_clc Specifying if a user received a reply based on an agent triggered action. An agent action is a predefined step within a subagent workflow.
            Escalation_Status_clc Specifying whether the escalation is a result of a session that moved to a different agent.
            Is_Session_Ended_clc Specifying if the session has ended.
            Is_User_clc Specifying if the entity is an agent or a user.
            Moment_Duration_clc The moment duration, in seconds.
            Quality_Score_clc A quality score, ranging from 1 (lowest) to 5 (highest). Indicates how relevant the agent's response was to the user's request.
            Session_Duration_clc The session duration, in seconds.

            Measures are quantitative values that can be aggregated, calculated, or compared. They represent the key performance indicators and metrics used to evaluate agent performance, such as counts, rates, averages, and percentages.

            Measures
            Field Name Description
            Abandonment_Rate_clc The abandonment rate percentage. Calculated by dividing the number of abandoned sessions by all sessions.
            Abandoned_Sessions_clc The number of sessions where the user left because of an error, browser crash, or time out.
            Average_Agent_Interaction_Latency_clc The average time it takes the agent to respond to the user's request, in milliseconds.
            Average_Interactions_Per_Session_clc The average number of interactions per session.
            Average_Moment_Duration_clc The average moment duration, in seconds.
            Average_Quality_Score_clc An average score, ranging from 1 (lowest) to 5 (highest). Indicates how relevant the agent's response was to the user's request across multiple sessions.
            Average_Session_Duration_clc The average session duration, in seconds.
            Average_User_Interactions_clc The average number of user interactions across all sessions.
            Deflection_Rate_clc The deflection rate percentage, calculated by dividing the number of deflected sessions by all sessions.
            Deflected_Sessions_clc The number of sessions ended by the user or otherwise, but not by escalation or abandonment.
            Engagement_Rate_clc The engagement rate percentage, calculated by dividing the number of engaged sessions by the number of sessions.
            Engaged_Sessions_clc The number of sessions where the user received a reply based on an agent triggered action. An agent action is a predefined step within a subagent workflow.
            Escalation_Rate_clc The escalation rate percentage, calculated by dividing the number of escalated sessions by all sessions.
            Escalated_Sessions_clc The number of sessions where an agent escalated a session to a human or a different agent.
            Quality_Score_Reasoning_clc The explanation for an assigned quality score.
            Success_Rate_clc The success rate calculated as the percentage of interactions where an agent's action-based response was provided, out of all interactions.
            Unique_Interactions_clc The number of distinct user and agent interactions recorded within a defined timeframe. An interaction consists of a user's request plus an agent's response.
            Unique_Moments_clc The number of distinct moments recorded within a defined time frame. Each moment aggregates multiple user and agent interactions centered on a specific request.
            Unique_Sessions_clc The number of distinct sessions recorded within a defined time frame. A session includes the entire exchange between a user and an agent.
            Unique_Tags_clc The number of unique tags created within a defined time frame.
            Unique_Users_clc The number of distinct users interacting with the agent within a defined time frame.
             
            Loading
            Salesforce Help | Article