The Deterministic Gap in Agentic Commerce
AI-generated image · US National Wire
Industry experts warn that probabilistic AI models clash with the precision required for financial transactions and e-commerce logistics.
The tech and financial sectors are increasingly focused on "agentic commerce," where AI agents handle shopping and payments on behalf of users, as first reported by The Register. According to a Mastercard report, ICSC and McKinsey project this could reach $1 trillion in US consumer retail revenue by 2030, while Gartner expects AI agents to intermediate over $15 trillion in B2B spending by 2028.
However, significant friction remains. The Register notes that Amazon has already blocked bots from Meta (Muse) and Perplexity. Beyond platform restrictions, Lindsay Walker, a product manager at Hedera AI Studio, told The AI Conference in San Francisco that agentic commerce is not yet a reality. Walker argues that current e-commerce infrastructure is built for humans and fails when applied to AI.
Specifically, Walker highlights the conflict between probabilistic AI models and the need for deterministic outcomes. Because money and value are too critical to be left to interpretation, she asserts that transactions require "deterministic gates" that cannot be bypassed by probabilistic models.
Functional hurdles also persist. Walker described attempts to use "agentic wallets" or cards that failed because the AI agents could not access DOM elements on web pages to enter credit card numbers or log into accounts to retrieve shipping addresses. Walker expects agentic commerce to become a reality within five years, provided the industry develops a mechanism to tether human identity to these agents.