The mParticle Agent helps you explore data in your workspace, build supported real-time audiences, and work with Predictions. It can recommend a prediction type for your business goal, create Future Behavior predictions, inspect and interpret supported existing predictions, and prepare supported audiences after you select an exact target range. Start with a question, problem, or business goal in plain language.
\nWhen your request leads to an action (like building an audience), the agent prepares a proposed configuration for you to review and approve rather than executing the change itself. Throughout the conversation, it explains its findings and assumptions and asks for clarification when needed. You remain in control by reviewing the Agent’s proposed configuration and explicitly approving the change before it is applied.
\nUse the mParticle Agent to:
\nWhen you submit a request, the mParticle Agent interprets your intent and uses the available context in your workspace, including Data Catalog metadata, aggregate metrics, and permission-gated data. It identifies relevant information, explains its findings, and asks for clarification when it needs more detail.
\nFor supported tasks that involve a change, the agent translates your request into technical criteria and settings and prepares a configuration for your review. You can refine the configuration through the conversation before approving it. The agent does not apply the change until you explicitly confirm it.
\nThe mParticle Agent can summarize workspace data, prepare and create one top-level real-time audience per creation flow, recommend Future Behavior, Similar Customers, or Best Offering for a business goal, and create Future Behavior predictions. It can inspect and interpret supported existing predictions. After you select an exact target range, it can build one individual audience from a Future Behavior or Similar Customers prediction. An audience can combine eligible existing Composable Audience membership with live behavioral criteria.
\nYou complete these steps in the standard product workflow:
\nThe Agent does not edit or delete saved resources or change workspace configuration. It asks for clarification when it cannot identify the information needed for a task and requires your explicit approval before creating an audience or Future Behavior prediction. See Build Audiences and Create Predictions for feature-specific steps and limits.
\nThe mParticle Agent runs under your authenticated mParticle session and follows your existing mParticle permissions. It can access only the workspaces, environments, and data that your user account can already access.
\nThe agent uses OpenAI to interpret your requests and generate responses; your prompt and the relevant workspace context are encrypted when sent to OpenAI and are processed under mParticle’s enterprise agreement with OpenAI. Your prompts and data are not used to train OpenAI’s models. mParticle does not use one customer’s conversations, workspace data, business context, or configurations to generate recommendations for other mParticle customers.
\nThe agent is not designed to expose personally identifiable information (PII) in your conversations. It won’t list stored email addresses, phone numbers, or raw identifiers, and it doesn’t support the export of PII. For more information about what the agent can and can’t access, how your data is secured, and where your data is processed, see Data and Privacy.
\n