AI can help you pick products, but do you dare let it spend money for you?

AI is functioning as a commercial discovery and referral layer well before it becomes an autonomous buyer, with McKinsey survey data showing 53% of respondents use AI for shopping but only about 9% of the broader sample complete purchases directly through AI tools. The distinction matters because discovery, checkout authorization and settlement each carry separate economics, and the article argues investment opportunity hinges on which transaction costs software coordination can lower and who controls distribution, authorization and fulfillment. Stablecoins enter the frame as one payment option for agents, alongside cards, instant bank transfers and corporate accounts, with adoption likely where both parties already hold compatible assets. Historical retail data show US e-commerce rose from 0.6% of sales in Q4 1999 to 17.1% in Q2 2026, a gradual shift interrupted by a 5.1-point jump in Q2 2020, illustrating that infrastructure and delivery economics, not technology alone, drive channel adoption. Payment rails can scale faster than retail channels, as Brazil's Pix has reached over 170 million users and India's UPI logged 241.6 billion transactions in FY2025-26, while Apple Pay took more than a decade to reach roughly a quarter of surveyed US online adults. Next to watch is whether agentic checkout converts referral traffic into delegated purchasing, and whether merchant acceptance, pre-authorization frameworks and settlement costs improve enough to move AI from research assistant to transacting buyer.