Banks and Retailers Clash Over AI Agents Blocking Them May Win Time, Not the Future

Market analysis

Major banks and retailers are already moving to limit or block third-party AI personal agents digital assistants that can shop, compare rates and even transact on behalf of customers but experts say that exclusionary strategies are unlikely to succeed over the long run.

JPMorgan Chase, Charles Schwab, Amazon, Neiman Marcus and Kohls have all taken steps to restrict agent access after the launch of Metas Muse and similar agentic AI tools. JPMorgan CEO Jamie Dimon recently said the bank blocks external AI agents from operating on behalf of retail customers because of security risks. Amazon, Kohls and Neiman Marcus have also explicitly restricted agentic access, while many other large retailers are effectively blocking shopping bots through older rules designed to stop web scrapers. An audit by Billions Network and AmanTrue of 51 large U.S. retailers found only 11 have a deliberate public policy toward AI agents; five explicitly restrict them and 37 block agents incidentally via legacy bot rules. None of the retailers audited currently offers a verified path for an agent to complete a purchase.

At the same time, other big players are taking the opposite approach. Walmart announced it will allow verified AI agents to shop on customers behalf, saying agent access will present the same prices and applicable offers a human shopper would see. The divergence underscores a broader strategic split in the industry: defend and restrict, or embrace and integrate.

Key detail

Security and fraud concerns are central to the debate. Reports of rogue AI agents accessing customer accounts have attracted political attention, and payments and fraud executives warn the risks are tangible: fraudsters can mimic agents to steal credentials, and even legitimate agents can behave unpredictably if permissions are unclear. Visas head of payment ecosystem risk and control, Michael Jabbara, highlighted scenarios where an agent takes unintended actions or a fraudulent agent compromises credentials problems that hinge on authenticating who an agent is and what authority it has.

Consulting and research data suggest customers want agent-driven services but demand control. Accentures research finds 65% of consumers are open to using a GPT-like financial assistant through a GenAI platform or digital wallet, and 71% would welcome an AI assistant embedded in their primary banks mobile app. Accentures Michael Abbott stresses that consumers want both convenience and control: 82% said they would want to approve every AI action and 79% want the ability to pause or stop AI activity immediately. Abbott argues that institutions that build secure ways for agents to operate will fare better than those that simply resist.

There are also economic incentives at play. Amazons internal Buy for Me features and its broader data control preferences show how incumbents can use agents selectively to preserve customer data and commercial advantages. Meanwhile, banks such as JPMorgan have begun charging data aggregators like Plaid for account access, passing costs to third-party apps and potentially to end consumers through subscription fees.

Technical and operational challenges add another layer. Experts warn of agentic flooding, where millions of agents generate high-frequency, automated requests that overwhelm sites or customer-service systems. Princetons Ed Felten notes that distinguishing agent-driven clicks from genuine human interactions will be difficult for remote sites, suggesting prolonged blocking may be futile. Abbott frames the risk as an infrastructure problem: the industry must build authentication, permissioning and machine-to-machine frameworks that support trusted agents at scale.

Not everyone expects the gates to open quickly. Former BlackRock executive Joseph Chalom argues incumbents control access points and will preserve those gates to protect business models, citing examples where agents would erode airline and banking margins by rebooking tickets or moving idle cash to higher-yield accounts. Yet many experts say market pressure and customer convenience will eventually push firms toward interoperable solutions.

The strategic tension will also test brand loyalty. Where consumers once stayed with a brand for reasons beyond price, AI agents will prioritize outcomes price, approval rates, speed and transparent terms over nostalgia or personal relationships. FICOs Nikhil Behl notes that agents care about consistent approvals and transparent pricing, not logos, which could reshape how companies retain customers.

In the near term, firms will split between those who restrict agents while building defensive infrastructure and those who pilot verified agent integrations. Longer term, the likely winners will be institutions that offer secure, trusted data and execution services that agents can rely on whether those institutions own the interface or operate behind the scenes. Blocking agents may buy incumbents time, but history and consumer demand suggest integration and secure participation are the more sustainable strategies.