The story behind the agent
The gap between goals and data
Marketers think in goals. Software thinks in events and fields. That gap—between marketing strategy and the data points required to execute it—is where time gets lost. Historically, building an audience often required knowing the right events, attributes, values, and lookback windows before you could get started. But marketers don't naturally think that way—they start with business goals. Audience Agent was built to bridge that gap, helping marketers translate goals into audience strategy grounded in their own data.
What we actually built
Our north star was simple: help marketers get from a goal to a real next step grounded in their actual data. Because every customer has unique data, we built an agent that reasons through each specific workspace. Rather than relying on generic recommendations, it understands the exact signals available in each customer's workspace.
We deliberately started with a focused audience-building experience rather than trying to make the agent do everything. Instead of layering chat on top of existing workflows, we rethought how marketers interact with the platform—starting with the highest-value use cases and expanding from there.
How marketers actually work with it
When showing the agent to marketers, we expected them to type specific instructions for the agent to build an audience. Instead, they asked broad questions: "What data do I have here?" or "Where are the gaps in our current audience strategy?"
Experienced users started treating the agent like a knowledgeable colleague to think out loud with.
Customers were able to begin asking useful questions with very little instruction. But the bigger surprise came later. We expected data discovery to be especially valuable for newer users. Instead, even experienced marketers consistently started by exploring their workspace—asking questions like "What data do I have?" or "What audiences should I be thinking about?" before jumping into execution.
For example, when a marketer asks how to improve campaign performance, the agent analyzes their specific workspace signals, explains its logic, and recommends an audience strategy. This shift is critical because it eliminates the technical friction that often stalls campaigns, empowering marketers to execute their strategy instantly without needing to manually map complex events or navigate intricate platform schemas.
Addressing the misconception: AI is a partner, not a replacement
One of the biggest misconceptions about agentic AI is that the goal is full automation. We don’t think that’s the right goal. The real value is helping marketers make better decisions faster by removing technical complexity—not removing the human. A well-designed agent knows its limits. It asks clarifying questions, explains its reasoning, and keeps the marketer in control.
What makes this different
What makes Audience Agent different isn't that it has a chat interface. It's that every recommendation is grounded in both the customer's data and the capabilities of the platform. Because mParticle has spent more than a decade helping brands unify customer data, the agent can reason over the events, attributes, predictions, and signals that actually exist in each workspace and recommend strategies the customer can actually execute. Instead of generic marketing advice, it helps marketers move directly from business goals to actionable audience strategies built around their own data.
Where we go from here
We’re looking forward to expanding the agent's ability to support more advanced audience workflows. We’re starting with audience creation, but we are actively building out advanced capabilities to support orchestration across our entire platform ecosystem. Next, we're expanding into predictions and deeper data exploration—helping marketers understand what data matters and what to do next. Ultimately, our vision is simple: marketers shouldn't have to learn the mechanics of a CDP before they can create value from it. They should be able to start with what they're trying to accomplish, and the platform should help them get there.
