AI Agent Generative AI Usage Data Model

About AiAgentGenerativeAiUsage

Each record represents a single generative AI interaction event processed through the AI Metering Service. The object captures the billing and metering decisions from AI services (for example, LLM Gateway and Agent Reasoner), including whether the interaction was billable, usage quantity, token counts, and identifiers for traceability and audit.

AI Agent Generative AI Usage Fields

The following table lists fields for the AI Agent Generative AI Usage DMO (AiAgentGenerativeAiUsage, with the developer name AiAgentGenerativeAiUsageDmo).

Query the Data Model

You can run ad hoc queries that pull relevant information from the DMO. Data 360 offers multiple ways to query your data. See Query Data in Data 360 to learn more.

Here are some example queries you can use.

Billable Usage by Agent

Use this query to identify the most expensive agents. It aggregates total usage and event counts, grouped by agent developer name and type.

SELECT
    AgentDeveloperName__c,
    AgentTypeCode__c,
    SUM(UsageQuantity__c) AS TotalUsage,
    COUNT(*) AS EventCount
FROM AiAgentGenerativeAiUsage_std__dlm
WHERE IsBillableIndicator__c = true
    AND Timestamp__c >= CURRENT_DATE - 30
GROUP BY AgentDeveloperName__c, AgentTypeCode__c
ORDER BY TotalUsage DESC

Token Consumption by Model

Use this query to identify which models consume the most tokens. It aggregates input, completion, total tokens, and event counts by model ID, provider, and class.

SELECT
    GenAiGatewayModelName__c,
    ModelProviderName__c,
    ModelClassType__c,
    SUM(PromptInputTokenCount__c) AS TotalInputTokens,
    SUM(PromptCompletionTokenCount__c) AS TotalCompletionTokens,
    SUM(PromptTotalTokenCount__c) AS TotalTokens,
    COUNT(*) AS EventCount
FROM AiAgentGenerativeAiUsage_std__dlm
WHERE Timestamp__c >= CURRENT_DATE - 30
GROUP BY GenAiGatewayModelName__c, ModelProviderName__c, ModelClassType__c
ORDER BY TotalTokens DESC
    

Usage by Channel

Use this query to compare usage across Voice and Messaging channels. It joins the generative AI usage and session tables to aggregate total usage and tokens by channel type.

SELECT
    s.ssot__AiAgentChannelType__c         AS ChannelType,
    s.ssot__RelatedMessagingSessionId__c  AS MessagingSessionId,
    s.ssot__RelatedVoiceCallId__c         AS VoiceCallId,
    u.AgentDeveloperName__c,
    SUM(u.UsageQuantity__c)              AS TotalUsage,
    SUM(u.PromptTotalTokenCount__c)      AS TotalTokens
FROM AiAgentGenerativeAiUsage_std__dlm u
    JOIN ssot__AiAgentSession__dlm s
        ON u.AiAgentSessionId__c = s.ssot__Id__c
WHERE u.Timestamp__c >= CURRENT_DATE - 30
GROUP BY s.ssot__AiAgentChannelType__c,
    s.ssot__RelatedMessagingSessionId__c,
    s.ssot__RelatedVoiceCallId__c,
    u.AgentDeveloperName__c
ORDER BY TotalUsage DESC
LIMIT 20

Build Data 360 Reports

To analyze Knowledge/RAG Quality Data and Metrics, build reports in Data 360 that pull relevant information from the AiAgentGenerativeAiUsage DMO. You can use calculated insights to create metrics that can trigger alerts. To learn more, see Data 360 Reports and Dashboards.