As autonomous artificial intelligence agents transition from experimental lab projects to functional tools for consumer convenience, the digital marketplace is facing an unprecedented operational shift. These agents—software programs designed to navigate the web, execute transactions, and manage logistics on behalf of human users—are effectively stepping into the role of a digital proxy. When an agent visits a website to perform a task, such as purchasing groceries, booking travel, or checking financial records, it interacts with backend systems designed for human behavior. This creates a fundamental friction point for businesses: how to authenticate, manage, and distinguish between a human customer and an automated delegate.
Throughout October, the commercial landscape witnessed three distinct approaches to this challenge, signaling a lack of industry consensus that will likely persist as agentic commerce continues to mature. While some organizations have moved to fortify their perimeters with total blocks, others have adopted a policy of caveat emptor, placing the security burden squarely on the user. Meanwhile, industry leaders like Walmart are pioneering more sophisticated, granular integration models that treat agents as distinct entities within the existing customer ecosystem.
The Spectrum of Response: Blocking, Liability, and Integration
The most restrictive response, adopted by institutions like Chase, involves the proactive blocking of AI agents. By identifying the non-human traffic patterns associated with automated scrapers and task-oriented agents, these institutions effectively bar the door. From a security and compliance standpoint, this is the path of least resistance. It eliminates the risk of bot-driven fraud, credential stuffing, and unexpected server strain. However, it also inherently limits the utility of the platform for power users who seek to leverage AI to streamline their financial management.
At the opposite end of the spectrum, other financial giants, including Bank of America, have opted for a "user-risk" model. In this framework, the bank permits the connection but shifts the responsibility of verification and potential misuse to the customer. If an AI agent performs an unauthorized transaction or interacts with sensitive data in a way that leads to a loss, the onus remains on the human account holder to remediate the situation. This approach acknowledges the reality of the evolving web but creates a significant barrier to mainstream adoption, as the average consumer is ill-equipped to audit the decision-making logic of an autonomous agent.
Walmart has emerged as a middle-ground innovator, proposing a strategy of "Agent Profiles." This method treats the AI agent as a secondary, authenticated identity linked to a primary customer account. By creating a separate profile, Walmart allows its systems to monitor the agent’s behavior—such as its browsing patterns, price sensitivity, and purchase history—without conflating that data with the human user’s direct activity. This distinction is crucial; an AI agent may browse significantly more inventory than a human, potentially skewing analytics. By segmenting this data, the business maintains a clearer view of its traffic while providing a sandbox where the agent can operate securely on the customer’s behalf.
The Mechanics of Agentic Commerce
The necessity for these varying strategies stems from the fundamental difference between human behavior and algorithmic operation. A human shopper has a finite capacity for browsing; an AI agent can scan thousands of pages, perform real-time price comparisons across multiple retailers, and execute complex order logic in milliseconds.
Walmart’s argument for the "Agent Profile" is rooted in the transparency of intent. As the company noted, current systems often force an agent to sign in using the customer’s credentials, effectively masking the automation. This leads to a blurring of lines where all purchases—whether human-initiated or bot-initiated—are aggregated into a single, undifferentiated bucket. By formalizing the agent as a linked profile, Walmart aims to create a system where limits can be applied to the agent (e.g., a "spending cap" or "access restriction") while the primary account remains the source of truth for the transaction.
This is analogous to a digital proxy; similar to how one might send an assistant or a family member to a physical store with a budget, the retailer requires a mechanism to verify that the person (or entity) acting at the register is authorized to do so. Without this verification, the "dog at the market" analogy—where a pet is sent to fetch goods—fails, because the merchant has no way to confirm who is paying or where the goods are going.
Data and Reality: The Scale of Current Agentic Activity
While the rise of AI agents is frequently discussed as a disruptive force, current volume metrics suggest that the industry is still in its infancy. Data from DoorDash, provided during their Q3 earnings call, serves as a vital reality check. CEO Tony Xu noted that the platform has handled "several hundred orders" via AI agents to date. When contextualized against the backdrop of 993 million orders processed by the company in the same quarter, it becomes clear that agentic commerce currently represents a statistically negligible, though strategically significant, portion of total traffic.
This scale provides a brief window of opportunity for businesses to build out their infrastructure. However, the growth trajectory is expected to accelerate as companies like OpenAI, Meta, and various startups integrate agentic capabilities directly into consumer-facing browsers and operating systems. The current "several hundred" orders will likely scale into the millions as these tools transition from niche tech-enthusiast utilities to standard features in everyday applications.
The Quest for Standardization: The Personal Agent Protocol
The ambiguity surrounding agent behavior has led to a collaborative effort to establish industry-wide standards. The Personal Agent Protocol, spearheaded by companies like Sierra and Meta, aims to provide a framework for how agents identify themselves to websites. As of November 1, more than 40 organizations were involved in the development of this protocol.
The protocol’s draft includes recommendations for agents to sign their requests using "Web Bot Auth," a security standard that allows an agent to cryptographically prove which entity is operating it. This is a critical development for website owners who need to filter "good" bots—those acting on behalf of legitimate users—from "bad" bots engaged in scraping or malicious activity. Meta’s "Muse" agent has reportedly been testing this signing process since mid-October, signaling that large-scale tech companies are moving toward a voluntary compliance model to ensure their agents are not blocked by standard bot-mitigation tools.
Implications for Future Strategy
For businesses operating in the digital space, the arrival of widespread agentic traffic requires a shift in how they view their web presence. The "blind" block, while effective, will eventually lead to lost revenue as more consumers move toward platforms that accommodate their digital delegates. To prepare for this shift, organizations should consider three strategic imperatives:
- Infrastructure Readiness: Audit current bot-management software. Many existing solutions are designed to treat all non-human traffic as a threat. Businesses need to implement logic that can distinguish between malicious scrapers and verified personal agents.
- Account Architecture: Consider the "Agent Profile" model. Providing a dedicated space for agents allows for granular security controls, such as daily spending limits or restricted access to sensitive account information, while still enabling the convenience that customers demand.
- Identity Verification: Monitor the development of the Personal Agent Protocol and similar standards. Being an early adopter of these authentication frameworks will ensure that your business can seamlessly serve the next generation of AI-enabled consumers without compromising on security.
The transition to agentic commerce is not merely a technological upgrade; it is a fundamental shift in the relationship between the merchant and the consumer. Just as the transition from brick-and-mortar retail to e-commerce forced a total rethink of logistics and supply chain, the shift from human-navigated websites to agent-navigated interfaces will force a total rethink of user authentication and data management.
Ultimately, the goal for businesses is to achieve a state of "agent-aware" operations. This requires a shift from reactive blocking to proactive identification. The organizations that succeed will be those that provide a secure, transparent environment where the agent can act on behalf of the customer, while the business retains the necessary control to prevent systemic risk. As the volume of agentic orders grows from the hundreds to the millions, the businesses that have established these protocols will find themselves at a distinct competitive advantage, having already solved the "who is at my door" problem while their competitors were still trying to figure out how to lock it.


