The Deterministic Gap: Why 'Agentic Commerce' is Stalling at the Checkout

The tech and financial sectors are currently enamored with the concept of "agentic commerce"—the idea of automated software executing purchases on behalf of consumers. As The Register first reported, this trend has surged since early 2025, fueled by optimistic projections from major firms. Projections from ICSC and McKinsey suggest U.S. consumer retail revenue from agentic commerce could hit $1 trillion by 2030, while Gartner forecasts AI agents will handle over $15 trillion in B2B spending by 2028.

However, from an operator's perspective, there is a glaring disconnect between these trillion-dollar forecasts and the actual mechanics of a digital checkout. The core of the issue is the friction between probabilistic AI and the deterministic requirements of financial transactions.

During a presentation at The AI Conference in San Francisco, Lindsay Walker, a product manager at Hedera AI Studio, noted that agentic commerce is not yet a reality. While some companies claim to have developed agent cards or wallets, Walker stated these are not yet functional. She recounted a personal attempt to use such a tool where the AI agent failed because it could not log into her account to access a shipping address, nor could it interact with the DOM element on the webpage required to enter a credit card number.

In my view, this is where the industry's hype hits a wall. We are attempting to layer probabilistic models—which operate on likelihoods and interpretations—over a commerce infrastructure designed for absolute precision. As Walker put it, the way e-commerce is currently built works for humans, but not for AI. She argued that because money and value are too critical, transactions cannot be run by probabilistic models; instead, they require "deterministic gates" that cannot be bypassed or left open to interpretation.

Beyond the technical hurdles of web elements and logins, there is a systemic battle over access. The Register reports that Amazon has blocked bots from Perplexity and Meta's Muse bot, citing terms of use restrictions. This suggests that the giants of e-commerce are not yet willing to let their storefronts be reduced to simple API calls for automated assistants hunting for the lowest price.

There are also significant trust and regulatory hurdles. AI models have struggled with reliability, and there is a lack of transparency regarding how agents make purchasing decisions or if vendors influence those choices. Furthermore, U.S. lawmakers are already scrutinizing personalized pricing driven by AI algorithms, which could create further regulatory friction.

While Mastercard recently published a report exploring the requirements to make agentic commerce workable, the path forward remains steep. Walker suggests that the industry needs a mechanism to tether a person's identity to an agent to solve the "card present" dilemma, as agents will never physically possess a card. While she expects agentic commerce to become a reality within five years, the current state of the technology suggests that the "agentic" part of the equation is still far from the "commerce" part.