A Snowplow executive responds to a Databricks/Scott Brinker report on the evolving martech stack, arguing that real-time AI agent decisioning requires a dedicated customer context layer built on structured behavioral event streams. The post distinguishes between customer records (who someone is) and customer context (what they're doing right now), contending that behavioral data must be schema-validated and identity-resolved at the point of collection before reaching the data platform. It outlines a four-stage agentic feedback loop (collect, resolve/enrich, serve, learn) and argues that the shift from analytics to real-time decisioning demands fundamentally different infrastructure — one that serves both low-latency and historical context simultaneously on open formats within the customer's own cloud environment.
Table of contents
The data platform is now the center of gravity for real-time decisioningWhat is the Customer Context Layer? And why real-time decisioning depends on itComposability was always the architecture, not the featureThe semantic layer begins before the platformThe Agentic Feedback Loop: How real-time decisioning actually closesSort: