AI just crossed an important threshold: for the first time, large language models can not only help people research products—they can complete the purchase in the same conversation. This emerging “agentic commerce” model is changing how discovery, comparison, and checkout happen online. It’s a shift that’s arriving faster than most expect, with real implications for how markets function and how consumers make decisions.
At the Conference Board’s AI Leadership Summit, I had the chance to speak with leaders across retail, payments, tech, and logistics who are now confronting this reality. Their questions were pragmatic: How do we deploy AI responsibly? How do we maintain trust? What new risks do we need to anticipate as more transactions get mediated by AI agents rather than websites?
Those risks are already measurable. In recent studies from Columbia Business School and Microsoft Research, AI shopping agents showed meaningful—and often unexpected—biases in how they ranked products, responded to layout changes, handled “badges,” or navigated large catalogs. Some models made systematically different recommendation decisions based on subtle labeling differences; others could be gamed by low-quality sellers responding quickly with misleading claims. The takeaway is clear: these systems are powerful, but not yet reliably aligned with the true interests of the people they represent.
This is where Consumer Reports’ work is both practical and relevant. With AskCR, we’ve spent the past year building an independent, AI-powered advisor grounded in our testing data and free from commercial influence. In the process, we’ve learned firsthand where general-purpose models excel, where they struggle, and what it takes to deliver recommendations that users can trust. Industry leaders at the Summit said this perspective is increasingly valuable as they experiment with their own AI-driven customer experiences.
Through our research collaboration with Stanford’s Digital Economy Lab, we’re also helping define early principles for “loyalty by design”—a framework for how AI agents should disclose incentives, manage conflicts, support portability of user preferences, and be evaluated against measurable consumer outcomes. Thaw are quickly becoming core design questions for every organization embracing agentic commerce.
The transition to AI-mediated shopping is underway. If we get this right, agentic commerce can broaden choice, strengthen competition, and reduce the cognitive load of decision-making. But it requires clear standards, trustworthy intermediaries, and systems designed to put consumer interests first. At CR, we’re committed to helping shape agentic commerce for the better—ensuring that as AI takes on more responsibility in the marketplace, it does so in ways that enhance, rather than erode, consumer welfare.
