Skip to Content
View all events

Agentic Commerce

Published by O'Reilly Media, Inc.

Intermediate content levelIntermediate

Build systems that let AI agents search, decide, and buy

Course Outcomes:

  • Explain the difference between AI features, AI assistants, and agentic commerce systems
  • Design agent-friendly workflows using APIs, MCP-style tools, and structured interfaces
  • Apply patterns such as mandates, approval boundaries, idempotency, and structured errors to commerce flows
  • Evaluate the trust, safety, and architectural trade-offs involved in letting agents search, decide, and transact

AI agents have moved into workflows that can search, compare, decide, and act. An especially important test case for these agents is commerce systems because they force teams to confront a harder question than “Can an agent call a tool?” The real question is whether a business is willing to let an agent complete user intent safely, predictably, and within clear boundaries. That requires more than MCP alone. It requires agent-friendly APIs, explicit authorization, transaction safety, structured errors, and a design that treats agents as real participants in the workflow rather than brittle screen-scrapers.

Author, teacher, and JavaOne Rock Star Ken Kousen uses MockHub—a working mock ticket marketplace built as a realistic AI test bed—to show you what agentic commerce looks like in practice. You’ll explore the progression from ordinary AI features (recommendations, price prediction, chatbot support) to agentic capabilities such as MCP tools, delegated authorization through mandates, and protocol-level interoperability through ACP-style flows. By the end of the course, you’ll understand not only how to let agents use a system, but how to build a system that can safely trust agents to act.

This live event is for you because...

  • You’re a software developer, architect, product manager, or engineering leader who wants to understand how AI agents interact with real transactional systems.
  • You want to move beyond chatbot demos and learn what it takes to make systems agent-friendly in practice.
  • You’re interested in emerging standards and patterns around MCP, authorization, and agent-mediated workflows.

Prerequisites

  • An understanding of software development and APIs
  • Familiarity with web applications and JSON/REST concepts
  • Basic knowledge of programmatic access to AI tools
  • Some exposure to LLMs, coding agents, or AI-assisted development (helpful but not required)

Recommended follow-up:

Schedule

The time frames are only estimates and may vary according to how the class is progressing.

Why agentic commerce matters (20 minutes)

  • Presentation: Why commerce is a hard and revealing domain for AI agents; why “MCP alone is not enough”; the difference between tool access and trustworthy action

AI features versus agentic systems (30 minutes)

  • Presentation: Recommendations, predictions, and chatbots; the progression from AI-in-app to agents with authority; why this distinction matters for architecture and product design

Agent-friendly architecture (40 minutes)

  • Presentation: APIs versus browser automation; OpenAPI, structured outputs, and discoverability; MCP tools as capability surfaces; why brittle scraping is not a strategy
  • Break

Authorization, mandates, and safety (40 minutes)

  • Presentation: The real trust problem—who authorized the agent?; delegated authority and scoped mandates; approval boundaries, reversibility, and auditability; functional correctness versus plausible output

Transaction design, and reliability (35 minutes)

  • Presentation: Idempotency keys; structured errors and next-action guidance; async flows, callbacks, and failure handling; why transactional systems demand more rigor than chat demos
  • Break

MockHub case study (40 minutes)

  • Presentation: Recommendations, price prediction, and chatbot features; MCP server- and tool-based workflows; mandate-backed agent purchases; what MockHub teaches about the future of commerce systems

Emerging standards and strategic implications (20 minutes)

  • Presentation: MCP-, ACP-, and AP2-style thinking; what businesses gain or lose by becoming agent-friendly; why this is a product and business strategy issue, not just a technical one

Wrap-up and Q&A (15 minutes)

Your Instructor

  • Ken Kousen

    Ken Kousen is the author of the Kotlin Cookbook (O'Reilly), Modern Java Recipes (O'Reilly), Gradle Recipes for Android (O’Reilly), and Making Java Groovy (Manning), as well as O’Reilly video courses in Android, Groovy, Gradle, advanced Java, and Spring. A JavaOne Rock Star, he’s a regular speaker on the No Fluff Just Stuff conference tour and has spoken at conferences all over the world. Through his company, Kousen I.T., Inc., he’s taught software development training courses to thousands of students.

    linkedinXlinksearch

Skill covered

Application Programming Interface (API)