Orchestrating streaming intents across product and growth surfaces

Modern products live across multiple surfaces—web, mobile, email, push notifications, in-app messages. Each surface has its own context, constraints, and user expectations. The challenge? Keeping experiences coherent as users move between them.

We built an intent orchestration layer that streams user signals across all surfaces in real-time. When a user abandons a cart on mobile, the web experience knows. When they engage with a feature on desktop, their mobile app reflects that context.

The architecture is event-driven. Every user action generates an intent signal—not just what they did, but what they're trying to accomplish. These signals flow through a central orchestrator that routes them to the right surfaces at the right times.

For example, when a user starts a checkout flow on mobile but doesn't complete it, we don't just send a generic 'complete your purchase' email. Instead, the orchestrator analyzes their intent, checks their behavior patterns, and routes a personalized message that acknowledges where they left off and offers relevant incentives.

The key insight: intents are more valuable than actions. A click is just a click, but understanding that a user is 'exploring pricing options' or 'comparing features' enables much smarter orchestration.

We use a combination of real-time event streaming (Kafka), intent classification (ML models), and routing logic (custom orchestration engine) to make this work at scale. The system processes millions of intent signals per day with sub-100ms latency.

The result? A 25% increase in cross-surface engagement and a 40% improvement in conversion rates for multi-step flows. Users feel like the product understands them, regardless of which surface they're using.