AI Customer: Agentic Commerce

From personalized search to market learning

6 chapters64 min total
Keywords
AI customeragentic commerceproduct curationweb agentspersonalizationagent-ready websitesmachine marketingmarket memorysearch costsconsumer welfare
Overview

AI customer systems act for people across discovery, evaluation, purchase, and service. The series traces how preference models, product evidence, agent-ready websites, and platform memory shape reliability, competition, and welfare, while keeping benchmark, prototype, and model-based limits distinct.

Learning outcomes
  • Reliable AI customer behavior depends on the full information path: preference representation, evidence, action, and feedback—not the model alone.
  • AgenticShop shows that finding a plausible product is easier than satisfying a particular shopper's criteria on open-web tasks.
  • In a controlled prototype with 150 runs per condition, the agent-ready site raised strict success from 49.3% to 89.3% and cut average steps 30.4%, but did not remove reasoning errors.
  • The market model suggests that cheaper search can expand learning, while selective screening without useful explanations can narrow competition and consumer welfare.
Keywords
AI customeragentic commerceproduct curationweb agentspersonalizationagent-ready websitesmachine marketingmarket memorysearch costsconsumer welfare

Chapters

Preview every chapter · Unlock full text, audio, and video from chapter 4

Plans
Ch.1-3 FreeCh.4-6 Premium

FAQ

What is an AI customer?

An AI customer is software that searches, compares, recommends, and may act on a person's behalf within delegated limits; it is an intermediary, not a human or independent consumer.

Why do AI customers miss the best product?

AI customers miss the best fit when representation, product attribution, current evidence, constraint handling, or recovery breaks; a fluent explanation cannot repair an invalid link or missing proof.

How can a website improve AI customer shopping?

A website can improve AI customer shopping by making product identity, selected variants, evidence, time status, and authorized actions explicit. A controlled prototype found higher strict success and fewer steps, but residual reasoning errors remained.

What are the market effects of agentic commerce?

Agentic commerce can lower search costs and expand market learning, but selective screening can narrow the competitive field and reduce consumer welfare under the market model's assumptions, especially when rejection reasons are not preserved.