Imagine an AI assistant that takes care of your shopping. You say: “I need toilet paper and detergent”, and he searches the internet, compares prices, analyzes reviews and finalizes the transaction. Without having to open a dozen browser tabs. It’s total autopilot.
This is the promise of agentic commerce: delegating purchasing decisions to intelligent algorithms. It seems inevitable, even obvious. But there are two main obstacles. First: the world’s two largest retail platforms are actively blocking this evolution. Second: most people don’t want to give up the purchasing process. For many, shopping is not a tedious task. It is a form of discovery and entertainment.
In August 2025, Amazon blocked access to several AI agents, including ChatGPT, Claude, Perplexity, and Google’s Project Mariner, with recent legal disputes against Perplexity for use of “cloaked” agents. Shopify, in turn, has developed infrastructure to “support” agentic commerce, but with strategic restrictions: agents cannot complete transactions alone and only work with Shop Pay, requiring human review on automated payments.
Continues after advertising
The message from both companies is clear: no one will interfere in our direct relationship with the customer. Still, disputes like these could force gradual integrations in the future, driven by open protocols or regulatory pressures.
Data is worth more than incremental sales
In 2024, Amazon earned $56.2 billion from advertising. In the second quarter of 2025, it was US$15.7 billion. This advertising revenue already competes with traditional retail margins. Selling products has become secondary; what really matters is understanding how you buy.
The fuel for this machine is behavioral data: every hesitation between two brands of coffee, every sneaker viewed multiple times, every cart abandoned because of high shipping. These traces power targeted ads, recommendation systems, and dynamic pricing.
Continues after advertising
If an external agent makes the purchase, Amazon loses this valuable trail. It only sees the final transaction, without the intermediate process. And it loses something even more critical: the ability to predict future behavior and calculate customer lifetime value. Without this data, the company’s billion-dollar valuation collapses.
Shopify follows the same logic, focusing on payments. Each transaction via Shop Pay generates insights into purchasing patterns, which today are moderately monetized, but tomorrow can support a scalable advertising business. Amazon blocks it in the discovery phase; Shopify, in the transaction. They are different tactics, but with the same objective: protecting the most valuable asset.
Shopping is a Pleasure — and a Question of Trust
Even if platforms allowed full access, agentic commerce would face a greater challenge: consumers do not want to delegate the majority of purchases. Who would trust an agent to choose clothes, decor items, gifts or books? In these categories, the selection process is an essential part of the value.
Continues after advertising
Hesitation between options is not a problem. It’s part of the experience. For many people, shopping is an expression of identity. Amazon’s infinite scroll is not a defect; it is an intentional feature.
In addition to pleasure, there are issues of privacy and trust. Consumers resist autonomous agents for fear of mistakes — such as wrong purchases or inappropriate recommendations, as seen in recent cases where AIs suggested incompatible products or ignored personal preferences. Data breaches are also a concern, especially with the rise of cyberattacks on AI platforms. This drastically restricts the agent trading market to repetitive and utilitarian items, such as basic replacements. Precisely boring, low-margin purchases that generate little valuable data. The niche is small and unprofitable, although tools like Amazon’s Rufus have already demonstrated success in generating additional sales in controlled scenarios.
Counterarguments are not convincing
Proponents of agentic trading make two main arguments. The first: “Amazon and Shopify are losing incremental sales.” True, but irrelevant. The marginal gain in volume does not compensate for the billion-dollar loss in market intelligence. It’s like exchanging gold for crumbs. The same mistake as aggressive promotions that inflate revenue, but destroy margins and brand value.
Continues after advertising
The second: “Consumers will adopt independent agents en masse.” This ignores the history of the internet. The open web has evolved into walled gardens like Facebook, Google and Amazon. Controlling the entire user journey always generates more value than being just a link in a fragmented chain.
Furthermore, Amazon already develops its own agents, such as Rufus and “Buy For Me”. If the agent is the future, it builds it internally, within its closed ecosystem, with exclusive access to the data that makes the difference.
Some suggest brands advertise directly to agents. But this model has fundamental flaws: if the agent receives payment for advertisements, they lose their neutrality; if it operates with a fixed user budget, it prioritizes only discounts, creating artificial behaviors. There is no sustainable incentive structure.
Continues after advertising
For an independent agent to compete, it is not enough to have a better search than Amazon. He needs to build a complete retail infrastructure: logistics, catalog, suppliers, payments and fulfillment. In short, it needs to become an Amazon. And replicate years of behavioral data that no money can buy. Recent examples, such as AI failures in automated purchasing (e.g., agents ignoring allergies or ethical preferences), highlight the risks and reinforce why adoption is slow.
Consolidation, not revolution
Agentic trading will become a reality, but not necessarily as an independent venture. The chance of an external agent capturing a relevant market share in the next five years is low, although not impossible — signs such as partnerships with tools like Stripe and open protocols suggest a gradual evolution. This goes against the current incentives of the largest retail platforms, but external pressures could accelerate changes.
What we will see is consolidation: the same companies that dominate e-commerce will lead agentic commerce. Amazon already has Rufus and “Buy For Me”; Shopify will follow suit. All operating in closed ecosystems, under proprietary rules and with exclusive data.
In the end, agentic commerce will not disintermediate anyone. On the contrary: it will deepen current intermediation. You will delegate choices to algorithms that follow the platform’s rules. It’s the same old story, repackaged with generative AI and sold as disruption.
Welcome to the future. It looks suspiciously like the present, only even more closed.
