Amazon's CEO on agentic commerce: what he got right, and where the critique falls apart
On Amazon's earnings call, its CEO fielded a key question about agentic commerce - how AI agents will help customers discover and buy products. To his credit, he acknowledged that agentic commerce has exciting long-term potential and could "expand the amount of shopping that happens online," and he revealed that Amazon is "having conversations with and expects over time to partner with third-party agents" - suggesting eventual collaborations with platforms like ChatGPT, Perplexity, or Claude.
The disconnect came next. He claimed third-party agents currently deliver a poor customer experience, citing no personalization, no shopping history, and inaccurate delivery estimates and pricing. Those criticisms do not match the current reality of the leading AI platforms: ChatGPT's memory feature has remembered user preferences and past searches since September 2024, the platform maintains order history and defaults to previously used parameters like shoe sizes, and pricing errors have become rare since the latest model generation.
On referral traffic, he dismissed agentic platforms as "a very small subset." The catch: Amazon is actively blocking bot traffic from AI platforms, which by design keeps that referral share near zero - a circular argument. And there is one development Amazon cannot block: AI browsers that run as the user rather than as an external bot. When the AI operates inside the customer's own browser session, blocking bots stops working entirely.
The read for sellers: there is a curious contradiction in Amazon's position - its CEO sees agentic commerce as the future while Amazon simultaneously walls off the platforms building it. As AI-powered shopping evolves, watch this space closely. With unblockable browsing technology now in play, Amazon's ability to control this channel is diminishing, and when it flips to partnering, the sellers already optimized for AI discovery will benefit first.
In other news from the same week: a new industry report ranked Amazon's global marketplaces by real seller opportunity rather than headline GMV, using seven dimensions including revenue, traffic, longevity, and activation. The United States leads clearly with $557K average revenue per seller and 60 percent of new sellers making their first sale within 12 months. Japan ranks only 8th in GMV but excels in seller longevity - 15.5 percent of sellers still active after five years versus 7.8 percent in the U.S. Germany is Europe's proving ground at $335K per seller; the UK is Europe's largest revenue opportunity at $342K but with post-Brexit activation challenges; Canada offers a conservative testing ground at $109K per seller. The lesson: the best marketplace depends on your strategy, not just market size.
Amazon has also completed the rollout of Custom Analytics, a unified platform consolidating more than 100 business metrics into a single customizable interface - account health, advertising results, reviews, and profit margins in one place, with heat maps, funnel diagrams, drag-and-drop report building, and template dashboards for quick setup. Sellers previously had to compile all of this from separate reports.
Quick hits: A major industry forecast predicts AI will influence 21 percent of purchases this holiday season, worth about $263 billion; Walmart debuted two Black Friday events; a Mastercard-PayPal agentic AI pact is shaking up checkout; and Amazon unveiled three new systems to speed fulfillment and delivery.
