NEW YORK – November 04, 2026 -- Hunting navigation app onX has cut the time needed to build predictive audience attributes from two weeks at best, and often closer to a month, to roughly one day, using Rokt's mParticle Agent.
Plain-language prompts replace external team coordination
Through the agent's Predictions capability, onX marketers describe the touchpoint they want to predict in plain language. The agent configures the prediction for review and approval, then makes the predictive attribute available for audience creation.
Previously, building a prediction required standing meetings with an external team and work to locate where data lived. Myka Arneson, Senior Marketing Technology Manager at onX, said the company had "zero self-service" before the rollout. She added that the agent surfaces patterns her team would not have thought to look for.
onX shifts from high-volume campaigns to targeted messaging
onX uses Rokt mParticle as its central platform for building audiences. With faster access to predictive attributes, its marketers are moving away from broad, high-volume campaigns toward communications based on individual member behavior and interests. According to the announcement, this reduces unnecessary marketing messages without sacrificing business performance. The source provides no quantified performance results.
Nick Craig, Head of Rokt mParticle GTM, said marketers can identify what they want to predict, create the prediction and activate it without a technical background.
Agent targets planning stage rather than content creation
Rokt contrasts the tool with much of the industry's agentic AI focus on content creation and campaign execution. mParticle Agent applies AI earlier, translating business objectives into predictive insights and actionable audiences.
The agent is the agentic layer of the Performance Engine in Rokt mParticle, the company's performance customer data platform for global enterprises. Marketers can also explore customer data and build audiences with plain-language instructions. All proposed logic is shown for review before it is saved, and activation runs through existing Rokt mParticle workflows.