Klaviyo Customer Agent: Building Bi-Directional Conversational Commerce at Scale

This presentation explores how Klaviyo is extending its core marketing philosophy — delivering the right message to the right person at the right time on the right channel — into fully bi-directional customer conversations through Customer Agent. Rather than treating engagement as a one-way outbound interaction, Customer Agent enables brands to continue those journeys naturally through real-time, personalized dialogue across channels.

A primary focus of the talk is how Customer Agent combines conversational AI with Klaviyo’s real-time customer and event data to create context-aware interactions that feel coherent, relevant, and actionable. We will examine how outbound campaigns, customer behaviors, product signals, and historical engagement are unified into a continuously evolving conversational context that allows brands to move seamlessly from marketing to interactive commerce and support experiences.

The presentation will also cover the high-level architecture behind Customer Agent and the engineering tradeoffs involved in building reliable conversational systems for commerce use cases. In particular, we will discuss how Klaviyo approaches determinism within an agentic system: balancing LLM flexibility with structured orchestration, retrieval constraints, policy enforcement, and deterministic business logic to ensure high-quality interactions every time.Topics include workflow orchestration, grounding and retrieval strategies, state management, tool invocation, evaluation pipelines, and safeguards that maintain consistency, trust, and brand alignment at scale.