Abstract
In 1998, Amazon.com invented item-to-item collaborative filtering —the algorithm behind “customers who bought this item also bought”— fundamentally changing how online retailers connect people with products. That innovation replaced earlier user-based approaches with a scalable, item-based similarity model and became one of the most influential recommender systems in the history of e-commerce. In the nearly three decades since, online retail personalization has evolved through successive waves of machine learning: from matrix factorization and contextual bandits, through deep learning ranking
models operating at massive scale, to today’s emerging inflection point — agentic AI.
This panel brings together technology and business leaders from Amazon and AWS to explore what the “agentic storefront” looks like: a retail experience where AI agents don’t just recommend, but reason, plan, and act — autonomously curating product discovery, orchestrating marketing campaigns, and adapting in real time to shifting customer intent. Beyond the technical frontier, the panel will examine the business implications of this shift: how agentic systems reshape customer engagement and lifetime value, what new economics emerge when AI agents mediate the relationship between brands and shoppers, and how organizations must evolve their strategies, metrics, and operating models to harness autonomous personalization at scale. Panelists will trace the arc from that foundational 1998 algorithm to the current frontier, bridging the MLresearch that enabled each leap with the commercial realities of deploying agentic systems in one of the world’s largest storefronts.
Target Audience
This panel targets researchers, applied scientists, and ML engineers working on recommender systems, personalization, or agentic AI, as well as business and product leaders in e-commerce and retail technology evaluating the impact of autonomous AI on customer engagement and merchandising strategy. Graduate students in AI/ML will also benefit.
Attendees should have a working familiarity with core machine learning concepts (supervised learning, collaborative filtering, neural networks) and a general understanding of how recommender systems operate in production. Prior exposure to reinforcement learning or large language models is helpful but not essential. No programming skills are required.
Presenters
- Carmen Nestares
- Kira Shabalin
- Alexandros Karatzoglou
Significance of the Topic, Timeliness, Novelty, Relevance to the AI Research Community
The online retail storefront is one of the most complex AI-driven systems in production today. It orchestrates product recommendations, promotional campaigns, marketing messaging, visual content, shopping themes, and deep customer understanding — from individual preferences to household composition — into a cohesive experience served to hundreds of millions of customers. Each of these surfaces has historically been powered by specialized ML models operating in relative isolation. The emergence of agentic AI creates the possibility of unifying these capabilities under autonomous systems that reason across the full storefront, coordinating what to show, how to message it, and when to adapt — in real time.
While agentic AI is generating significant research interest, most academic work focuses on single-task agents or conversational assistants. The challenge of coordinating across the full storefront — simultaneously curating product discovery, generating promotional content, selecting imagery, and tailoring messaging based on rich customer context — represents a fundamentally different problem at a different scale. This panel addresses that gap with an end-to-end perspective that examines not just the technical coordination, but also the implications for business strategy, customer engagement models, and organizational design — an interdisciplinary lens spanning ML research, systems engineering, and commercial strategy that is rarely represented in a single session.
The topic aligns directly with two IEEE CAI verticals: “Generative AI Models, AI in Education and Agentic AI” and “Business Intelligence / Finance.” It will appeal to researchers exploring agentic architectures and multi-task coordination, practitioners building large-scale ML systems, and business-minded attendees evaluating how autonomous AI reshapes the economics of digital retail.