Abstract

Agentic AI promises systems that don’t just respond, but plan, decide, and act. Yet for many e-commerce teams, the concept remains abstract - and often confused with LLM-powered chatbots or tool integrations.

This hands-on workshop brings a practitioner’s perspective from the automotive domain, where conversational and voice-based AI systems have been deployed in safety-critical, real-world environments. Participants will explore how in-car voice assistants evolved from rule-based commands to LLM-powered systems - and shares challenges that have been faced along the way and lessons-learned.

Using concrete automotive examples as a “hard mode” reference, the workshop translates these learnings into the e-commerce context. Attendees will learn how to distinguish assistants, agents, and truly agentic systems; where autonomy creates value; and where it must be constrained.

The session combine a short impulse talk with group exercises. Participants will design their own agentic commerce use cases, define autonomy boundaries, and apply guardrails inspired by automotive AI systems.

By the end of the workshop, attendees will have a clear mental model, practical tools, and a realistic roadmap for moving from chatbots to agentic AI in e-commerce - without hype, and without losing control.