Topic

Agentic Commerce: Causal Modeling Practices in Commercial World Models

Agentic AI is driving commercial systems from prediction-driven paradigms toward decision-driven intelligence. Yet the core challenge lies not in building agents themselves, but in constructing a commercial world model that accurately captures the relationships among decisions, environment, and outcomes. Within the Agentic Commerce framework, buyer agents, seller agents, and platform agents jointly participate in market interactions, transforming business decision-making into a dynamic multi-agent, multi-strategy system. In this talk, drawing on practical experience in retail pricing, we will introduce our foundational exploration on the Sell-side Agent. By leveraging large language models as semantic priors and integrating temporal dynamic modeling with causal inference methods, we move beyond traditional forecasting toward counterfactual modeling—estimating demand responses to price changes. This approach provides the foundational capabilities required to build truly decision-oriented commercial world models. Outline: Agentic Commerce: The Next Paradigm of Commercial AI Commercial Systems from a Multi-Agent Perspective: Buy-side / Sell-side / Platform Agents Decision Spaces and Environment Interaction: The Dynamic Complexity of Commercial Systems From Forecasting to Counterfactual Modeling Foundations of the World Model for Sell-side Agents LLM-Driven Semantic Priors and Causal Demand Modeling in Practice