The Agentic Commerce Delusion: Why Your Store Isn't Ready for Shopping Bots
AI-generated image · US National Wire
Industry hype suggests a trillion-dollar future for AI agents, but basic technical friction and probabilistic errors make the 'shopping bot' a distant reality.
The tech and financial sectors are currently enamored with the concept of "agentic commerce"—the idea of automated software making purchases on behalf of consumers. While the vision is seductive, as The Register first reported, the reality is that most retail infrastructure is fundamentally incompatible with AI agents.
According to reporting from The Register, the scale of this opportunity is massive on paper. A Mastercard report highlights projections from ICSC and McKinsey suggesting that US consumer retail revenue from agentic commerce could hit $1 trillion by 2030. Additionally, Gartner predicts that by 2028, AI agents will facilitate over $15 trillion in B2B expenditures. However, these figures are speculative, and the actual implementation is facing significant headwinds.
**Opinion: The Infrastructure Gap** From an operator's perspective, the industry is focusing on the wrong end of the transaction. We are obsessed with the "payment moment," but the actual commerce experience is broken for non-human users. As Lindsay Walker, product manager at Hedera AI Studio, noted at The AI Conference in San Francisco, the way e-commerce is currently built works for humans but fails for AI. We are trying to plug probabilistic AI models into a system that requires deterministic certainty.
**The Technical Friction** Retailers are already fighting the bots they are supposedly trying to court. The Register reports that Amazon has blocked Meta's Muse bot and previously blocked Perplexity's AI bot, with the latter resulting in a legal battle over a preliminary injunction.
Beyond the corporate warfare, the functional failures are glaring. Walker shared an example of a failed checkout attempt where, despite having a virtual card, the AI agent was unable to complete the purchase because it could not log in to access a shipping address or interact with the specific DOM element on the webpage required to enter credit card details.
**The Trust and Logic Problem** Even if the technical hurdles are cleared, the logic of agentic commerce remains flawed. Walker argues that while AI can assist with product selection, delegating full authority to an agent is where the system breaks down. Because money and value are too critical to be left to interpretation, commerce requires "deterministic gates" that probabilistic models cannot bypass.
Furthermore, there are systemic trust issues. AI models lack transparency regarding how purchases are made or if vendors are influencing decisions. This is particularly risky given that US lawmakers are already scrutinizing the use of AI algorithms for personalized pricing.
For now, the "agentic wallet" is more myth than reality. Walker expects it may take five years before agentic commerce becomes a functional reality. Until retailers stop treating their storefronts as human-only interfaces and start solving the underlying data and access problems, the shopping bot will remain a boardroom fantasy.