The Funnel Reality
What an agentic commerce system can do and what a shopper will allow it to do are no longer a matter of speculation. The gap is measurable. According to the Marketplace Shopping Behavior Report 2026, which surveyed 4,500 shoppers across five markets including the U.S. in October 2025, the consumer funnel for AI-mediated purchasing is narrow. While 58% of shoppers use AI tools to research products, only 37% have started a purchase journey through an assistant. When it comes to the final step—the actual transaction—that number drops to 17% who feel comfortable completing a purchase through AI. Most tellingly, 76% of consumers demand a human review and approval step before any money changes hands.
The October Sprint
While consumers remain cautious, the vendor side has spent October in a state of high-velocity deployment. This sprint follows the Trust Paradox, a thesis explored here three weeks ago, which identified that infrastructure was sprinting ahead of actual readiness. October confirms the industry is building for a future the consumer has not yet inhabited.
The timeline of this push is precise. Mastercard initiated the cycle on September 30 with its agent intelligence services announcement. October saw a rapid succession of infrastructure launches: Cequence shipped its Agent Trust registry on October 7, the same day Cisco Webex announced its agentic AI framework. By October 13, Forter launched its Agentic Commerce Engine, and Harbor Insurance emerged from stealth. Two days later, PayPal and Glance announced an agentic commerce integration. The pilot phase is effectively over, as evidenced by the Money20/20 agenda, which has moved beyond experimentation into production-focused programming.
Cisco’s own survey from March 2026 found that 85% of organizations are experimenting with AI agents, yet only 5% are running them in broad production deployment. This leak between experiment and production mirrors the same structural gap found in the consumer funnel.
Corroborating the Gap
The data from the ChannelEngine report is not an outlier; it is part of a consistent pattern. IBM Institute for Business Value research from January 2026 found that while 45% of consumers use AI for parts of the buying journey, their usage is largely confined to research and deal-hunting. Attentive reports that only 10% of purchasers complete their transaction inside an AI tool, with 60% preferring to finish on a brand’s native site. Even when consumers express interest, as noted in the BigCommerce and PayPal research, the demand for human oversight remains a persistent friction point. The RSM synthesis of Riskified data shows that comfort levels with AI-led purchasing have actually declined over the past year, falling from 70% to 45%.
The Liability Wall
The supply gap sits on top of a harder disagreement: who carries the risk when the agent gets it wrong. Merchants are not eager to hand over the keys to their inventory. According to vendor-reported data cited by PYMNTS and Visa, 93% of merchants believe the AI provider should bear the financial loss for incorrect purchases, yet only 28% are willing to expose their full inventory to agents. This is the economic reality of the holiday peak. Forter reports that agentic orders have quadrupled since August, and automated fraud has doubled in the same period. The Q4 stress test arrives in weeks. The technology is ready. The bill for its mistakes is unassigned.