Automated AI agents that make purchases on behalf of consumers won't become a reality for at least five years, despite widespread enthusiasm from the tech and payments industry for so-called "agentic commerce," according to a report published Wednesday by The Register. The article covers a presentation by Lindsay Walker, product manager at Hedera AI Studio, at The AI Conference in San Francisco, where she outlined the technical barriers preventing shopping bots from functioning despite industry projections. While payment firms like Mastercard published forecasts this week suggesting the technology could transform retail, Walker and industry observers say fundamental infrastructure problems remain unsolved.
The report notes that Mastercard's new analysis cites ICSC and McKinsey projections suggesting agentic commerce in US consumer retail could generate $1 trillion in revenue by 2030, while Gartner anticipates AI agents could handle more than $15 trillion in business-to-business spending by 2028. However, real-world tests reveal the technology isn't operational yet. Walker recounted an example where an AI agent failed to complete a checkout process because it couldn't log into her account to retrieve shipping information and couldn't access the webpage element where credit card details should be entered, forcing manual intervention.
"Agentic commerce is not a reality yet," Walker stated during her presentation. "There are a lot of products out there claiming that they're creating agent wallets or agent cards. But when it comes down to it, it's not functional yet and hasn't truly emerged." She explained that probabilistic AI models can't be trusted with financial decisions, saying "Money is too important. Funds are too important. Value is too important. You need to have deterministic gates that can't be bypassed by a probabilistic model."
The analysis identifies several obstacles blocking adoption. Major e-commerce platforms like Amazon have already blocked automated shopping assistants from Meta and Perplexity, citing terms of use violations, and current websites calculate taxes and shipping based on customer location data that changes when agents operate from remote servers. Walker emphasized that the entire e-commerce infrastructure needs rebuilding so agents can interact at "machine speed," meaning payments, agreements, and proofs must move in fractions of a second with fractions of a cent per transaction. The x402 protocol for internet payments is under development but only addresses the payment moment itself, not product discovery, negotiation, settlement, or post-purchase logistics.
The report concludes that consumers currently experience only "agent-assisted shopping" where humans remain present at the moment of purchase and make all decisions, rather than autonomous commerce. Walker stressed that trust mechanisms linking a person's identity to an agent are essential before delegation becomes viable, particularly since payment cards are never physically present with AI agents. Getting today's e-commerce companies to accept frictionless transactions earning fractions of a cent instead of their current margins won't be easy, the report notes. The challenge isn't just technical capability but convincing an entire industry to restructure how it captures value, especially when regulatory scrutiny of AI-driven personalized pricing is already intensifying and consumer trust in opaque algorithmic decision-making remains fragile.