Back

Coach, Kate Spade and Kohl’s Restrict AI Agents From Completing Purchases

Fintech Digital Marketing

Blog

Oct 28, 2026

5 views

Technology

Coach, Kate Spade and Kohl’s Restrict AI Agents From Completing Purchases

AI shopping agents promise to make online shopping faster. They can search for products, compare options, and potentially complete purchases while customers do something else. But a recent incident involving luxury resale platform TheRealReal reveals a less obvious problem. An AI agent does not need to hack a website to disrupt it. Sometimes, simply using a legitimate feature faster and more persistently than a human can create trouble for everyone.

The Shopping Cart Becomes a Battleground

TheRealReal has reportedly strengthened its defenses against third-party AI shopping agents after a test revealed how automated systems could exploit its shopping-cart process.

The platform allows shoppers to place an item on hold for 20 minutes by adding it to their cart. According to reporting by The Verge, its journalists used Meta’s Muse shopping agents to keep items reserved for hours, preventing other customers from purchasing them.

TheRealReal subsequently said it would make it harder for shopping agents to log in, add products to carts, and complete checkout.

The issue is not merely automated traffic. It is the difference between a human shopper, who may browse casually and abandon a cart, and an agent that can repeat actions around the clock. A feature designed to give customers time to decide can become a tool for tying up inventory.

AI Shopping Is Growing Faster Than Retail Rules

Meta’s Muse agents have attracted substantial early interest. CEO Mark Zuckerberg said on the company’s earnings call that more than 10 million people in the United States and Canada had set up personal agents since their launch early the previous month. Agentic shopping was the dominant use case, according to his remarks.

Zuckerberg also said that more than a quarter of Muse users made an average purchase exceeding $200.

Those figures suggest that AI-assisted shopping is not simply a technology demonstration. Consumers are exploring ways to delegate product discovery and purchasing decisions to software.

But adoption creates pressure on the businesses receiving those agents. Retail platforms were built around human behavior, conventional browsers, and familiar checkout journeys. Autonomous software introduces different patterns of activity, including rapid requests, repeated cart updates, and continuous monitoring.

A system that works well for a person may not work safely when thousands of automated agents interact with it at machine speed.

Retailers Are Drawing Their Own Boundaries

TheRealReal is part of a wider group of businesses limiting third-party shopping automation.

According to the supplied reports, Tapestry, the parent company of Coach and Kate Spade, and department store Kohl’s do not allow AI shopping agents to complete purchases on their platforms. Online pet retailer Chewy and department store chain Neiman Marcus block some agents.

Delta Air Lines has also said it does not yet allow AI agents to buy tickets on behalf of travelers. Users have reported difficulties using Muse to shop autonomously on brand websites, including Adidas.

Amazon moved quickly to block Muse after Meta launched the tool, while offering its own shopping assistant, Alexa for Shopping.

The pattern reveals a key distinction in platform strategy. Many businesses are not rejecting AI as a category. They are deciding who gets to control the AI interface between the customer and the transaction.

A retailer may welcome automation built into its own platform while restricting an external agent whose behavior it cannot fully manage.

The Real Risk Goes Beyond Security

Much of the debate about AI shopping focuses on account takeovers, scams, data breaches, and unauthorized payments. These are serious concerns because an agent may need access to personal information, shopping histories, payment methods, and customer accounts.

Financial institutions have warned about the safety and privacy risks of agentic commerce. Banks are reportedly taking steps to block some third-party agents from accessing customer accounts because of concerns that rogue software could initiate unauthorized transactions.

Yet the incident at TheRealReal highlights a more immediate operational risk: legitimate features can behave differently when used autonomously.

An agent can reserve inventory without intending to purchase it. It can repeatedly check availability, place items in carts, or trigger systems designed for ordinary browsing. Even without malicious intent, automated behavior can create friction for other customers.

The result may be slower shopping, unavailable products, more abandoned transactions, and additional costs for retailers that must defend their systems.

Retail Platforms Want to Keep Control

For an e-commerce business, the customer relationship is a valuable asset. The platform controls product presentation, recommendations, promotions, checkout, and the data generated along the way.

Third-party agents can shift some of that control elsewhere. Instead of visiting a retailer’s website and seeing its curated experience, customers may ask an external assistant to compare products across multiple stores and select an option for them.

That could make shopping more convenient. It could also weaken the retailer’s ability to influence the decision through merchandising, brand storytelling, loyalty programs, and personalized offers.

The incentives are therefore mixed. Retailers may benefit when agents bring qualified buyers, but they may lose control over how customers discover products and complete transactions.

Amazon’s approach illustrates this tension. Blocking a competitor’s agent while offering its own assistant allows the company to experiment with AI shopping without giving up control of its own customer experience.

A New Layer of Competition in E-Commerce

The emerging contest is not just between retailers. It is also between the platforms that want to become the customer’s primary shopping assistant.

If an AI agent becomes the first place a consumer searches for a product, the agent may influence which retailers receive attention and which products are considered. That could make agent visibility a new competitive concern, alongside search rankings, advertising, and marketplace placement.

However, agents need reliable access to product information and purchasing systems to deliver on that promise. Restrictions, inconsistent integrations, and security controls can limit what they accomplish.

Retailers, meanwhile, need ways to distinguish useful automation from activity that harms the shopping experience. Blanket restrictions may be easier to enforce, but they could also exclude legitimate customers who prefer AI-assisted browsing.

The eventual balance will depend on whether retailers and agent providers can agree on secure, predictable ways for software to search, reserve, and buy products.

Practical Lessons for Smaller Retailers

Small and medium-sized businesses do not need to build an AI shopping system immediately. They do need to prepare for automated traffic and delegated purchasing.

First, review features that assume a human is acting at a normal pace. Cart reservations, limited-stock offers, login attempts, and checkout flows may need rate limits or additional verification.

Second, separate product discovery from purchasing authority. Allowing an agent to read public product information is not the same as allowing it to reserve inventory or spend a customer's money.

Third, monitor unusual activity, including repeated cart holds, rapid requests, and abnormal checkout patterns. Security measures should protect genuine shoppers without making ordinary purchases unnecessarily difficult.

Finally, keep product information accurate and consistent. As AI assistants become another way customers discover products, clear descriptions, pricing, availability, and return policies will help both human buyers and automated systems make better decisions.

The Next Shopping Interface Will Need New Rules

AI agents could make e-commerce more convenient by reducing the time spent searching, comparing, and purchasing. Meta’s early adoption figures suggest meaningful consumer interest. But convenience for the buyer does not automatically translate into a better experience for every retailer or shopper.

TheRealReal’s response demonstrates that the challenge is not limited to hypothetical cyberattacks. Automated agents can disrupt everyday commercial processes simply by operating at a scale and speed that existing systems were not designed to handle.

The next stage of AI shopping will depend on trust, technical safeguards, and clear boundaries around what agents can do. Retailers want customers to benefit from automation, but they also need to protect inventory, transactions, and the customer relationship.

The winners will be the businesses that make AI shopping useful without surrendering control of the systems that keep commerce working.