AI-Driven Shopping: Beyond the Shopping Cart
We are on the verge of a revolutionary change that goes far beyond simple automation or more intelligent chatbots. We are entering the era of agentic commerce.
TABLE OF CONTENTS
AI-Driven Shopping: Beyond the Shopping Cart
For decades, the rhythm of the online shopping journey has been a familiar one: you browse, click, add to a cart, and check out. It's a process we've mastered, a digital translation of the physical shopping trip. But this entire model—the very foundation of e-commerce—is on the verge of a revolutionary change that goes far beyond simple automation or smarter chatbots.
We are entering the era of agentic commerce.
Having spent over 20 years watching digital marketing evolve, I can tell you this shift feels different. It's not incremental. It's not just another algorithm update to prepare for. When I first saw the traffic data from generative AI sources during the 2024 holiday shopping season, I knew we were witnessing something fundamental. Adobe Analytics reported that traffic from generative AI sources increased by 1,300% year-over-year during the holiday season—and on Cyber Monday alone, that number hit 1,950%.
This new paradigm isn't about personalized recommendations; it's about empowering an autonomous AI agent to act as your personal procurement specialist. Imagine stating an intent—"Find me the best eco-friendly, waterproof trail running shoes under $100 for a trip next month"—and having an agent not only research options but also negotiate terms, coordinate logistics, and execute the purchase, all without your direct involvement in every step.
This shift from an active participant to a high-level goal-setter brings a host of surprising, counter-intuitive, and deeply impactful changes that extend far beyond mere convenience. It will fundamentally reshape the relationship between consumers, brands, and the very infrastructure of digital markets.
It's Not Just a Smarter Chatbot. It's a Fundamental Power Shift
The most critical change in agentic commerce isn't a better recommendation engine or a more conversational interface. The core evolution is the transfer of decision-making authority from the human to the AI. This is a profound departure from the e-commerce model we know.
In traditional online shopping, the human controls every single step. We browse, we filter, we compare, and we make the final click to purchase. With agentic AI, the user simply states an intent or a goal. The AI agent then autonomously plans, reasons, and executes the purchase. It's the difference between a tool that assists you and an agent that acts for you.
I've been studying this shift closely through my work, building AI-powered shopping assistants and brand visibility analysis tools. What I've observed is that we're not just changing the interface as we're fundamentally changing the shopping journey itself.
As Katherine Black, a partner in Kearney's Consumer and Retail practice, put it: "Agentic commerce stands to disintermediate shopping in a way we haven't seen since the dawn of ecommerce."
The numbers back this up. According to Adobe's latest research, AI-driven traffic to U.S. retail sites grew 4,700% year-over-year by July 2025.
And here's what's particularly significant:shoppers arriving from generative AI sources are 16% more likely to convert than those from traditional channels, leading to increased customer satisfaction. They're also 13.6% more engaged, with 44% longer visits and a 31% lower bounce rate.
This is impactful because it fundamentally redefines the role of the consumer. We are no longer digital window shoppers or cart-fillers. Instead, we become high-level goal-setters, delegating the complex, tactical execution of our needs to an algorithm that acts as our personal chief purchasing officer.
Your Brand Loyalty Might Become Obsolete
One of the most surprising and counter-intuitive consequences of agentic commerce is the direct threat it poses to traditional branding and marketing. The emotional connection and loyalty that brands have spent decades and billions of dollars building may become far less relevant.
AI agents are not designed to be swayed by clever advertising or aspirational messaging. They are designed to optimize a purchase by performing a complex, multi-criteria analysis and making strategic trade-offs a human might not. An agent won't just look at price and features; it will weigh them against shipping time, sustainability scores, delivery consistency, and user reviews, making reasoned judgments about which product combination provides the optimal outcome for a given intent.
This isn't speculation. Boston Consulting Group's research on agentic commerce makes the risks clear: "diminished direct access to customers, weaker brand loyalty, and a growing dependence on intermediary platforms." BCG found that AI agents prioritize price, user ratings, delivery speed, and real-time inventory over brand familiarity or loyalty.
The data is stark. A Kearney study found that six in ten U.S. consumers expect to use AI shopping assistants within the next year. But here's what keeps me up at night as a strategist: their modeling suggests potential EBIT erosion of up to 500 basis points for retailers, stemming from price transparency, smaller orders, and agent platform fees.
As Black noted in that same research: "The next retail arms race won't be for clicks or shelf space. It will be for algorithmic preference and being the brand an AI agent consistently selects."
In a marketplace where algorithms are the primary "shoppers," brands that cannot demonstrate a clear, data-backed advantage across a matrix of criteria risk being systematically ignored. Survival will depend not on emotional appeal, but on the ability to win a multi-variable, logical argument against every competitor, every single time.
Get Ready for the "Liability Storm"
While agentic commerce promises unprecedented convenience, it simultaneously opens up a massive and largely unresolved legal and financial risk. This new model creates a profound ambiguity around accountability that the current commercial ecosystem is not prepared to handle.
When an autonomous AI agent makes a mistake, for example, buys the wrong item, misunderstands a complex request, or purchases a product that fails to meet expectations, then, who is at fault and who pays for it?
Is it the consumer who set the initial goal, the technology company that developed the AI, the merchant who accepted the automated transaction, or the payment processor who facilitated it?
I've spent considerable time researching this question because it's fundamental to how brands should prepare.
The answer, according to Kris Nagel, CEO of fraud platform Sift, is sobering: "Missing from the conversation is who pays the bill when these agents make mistakes or are manipulated."
This creates what experts are calling a "liability gap." According to a recent panel at ChargebackX 2025, the consensus among fraud experts is that merchants will bear the brunt of these costs.
As Jamie George, VP at Ravelin, put it bluntly: "The card schemes aren't going to take liability. Neither are the customers, their issuers, or the AI model, who technically hasn't made any money. That doesn't leave many options—it's going to be the merchant."
The friendly fraud amplification is particularly concerning. Chargeback fraud already costs merchants an estimated $50 billion annually.
Riskified's global study found that friendly fraud chargebacks already comprise 75% of all disputes—and AI agents will provide bad actors with lucrative new ways to dispute legitimate transactions.
Here's the kicker: The Fair Credit Billing Act of 1974 limits consumer liability for unauthorized credit card use to $50 and provides chargeback rights for "billing errors" or "goods not as described."
But does an agent's misinterpretation constitute a billing error? Courts haven't decided yet.
This unresolved issue of accountability is not a minor detail; it is a major barrier to widespread trust and adoption. Without clear frameworks for assigning responsibility, the entire system risks being undermined by financial disputes and a lack of consumer confidence.
The Real Revolution Is a Stack of Boring-Sounding Protocols
The magic behind an AI agent seamlessly buying a product isn't a single technology. It's a layered "stack" of open, standardized protocols that must work together, each solving a different piece of the puzzle. This shared language allows any agent to talk to any merchant, preventing the need for millions of custom integrations.
In my work building AI tools and integrations, I've come to appreciate just how important this infrastructure layer is. It's not glamorous, but it's everything. Here is how the key protocols work in sequence:
- Model Context Protocol (MCP): The Agent's Memory. First, an agent needs memory. Developed by Anthropic and released in November 2024, MCP provides the context, allowing an agent to remember its goals and past actions as it moves between different systems. Think of it as the USB-C of AI—a universal connector that standardizes how AI systems access data and tools. The adoption has been remarkable: OpenAI officially adopted MCP in March 2025, followed by Google DeepMind confirming support in April 2025.
- Agent-to-Agent Protocol (A2A): The Communication Layer. Next, agents need to talk to each other. Google's A2A protocol allows a buyer's agent to coordinate and negotiate with a merchant's agent, comparing options or clarifying details without human intervention.
- Agentic Commerce Protocol (ACP): The Transactional Workflow. With context and communication established, ACP manages the actual shopping process. OpenAI and Stripe co-developed ACP and launched it in September 2025 with "Instant Checkout" in ChatGPT. It standardizes how an agent performs the core functions: searching for products, adding items to a cart, and initiating the checkout. Currently, U.S. ChatGPT users can buy from Etsy sellers directly in chat, with over a million Shopify merchants coming online soon.
- Agent Payments Protocol (AP2): The Security & Authorization Layer. Finally, and most critically, the transaction must be secured. Google launched AP2 in September 2025 with backing from more than 60 partners including Mastercard, American Express, PayPal, Coinbase, and Salesforce. AP2 provides this layer by using cryptographically signed "Mandates"—tamper-proof digital contracts that create a verifiable audit trail proving a user actually authorized the agent to make a specific purchase.
What's fascinating is how quickly this infrastructure is being built. PayPal announced in October 2025 that it will adopt the Agentic Commerce Protocol, connecting its global merchant network to OpenAI and creating a platform for tens of millions of small businesses to sell within ChatGPT.
Forget Website Design. Your Next Big Project Is a Data Cleanup
For years, e-commerce success has been tied to optimizing the human experience: intuitive website design, compelling product photography, and persuasive marketing copy. In the agentic era, this focus undergoes a radical shift. A company's most important asset will no longer be its user-friendly website, but the quality and completeness of its structured product data.
This is something I've been talking with clients for months now. AI agents do not "see" a website; they read its underlying data. Agents rely entirely on complete, accurate, and machine-readable product metadata to make purchasing decisions.
If a product's specifications are incomplete, its compatibility information is missing, or its attributes are described in vague marketing terms, the agent will either ignore the product entirely or make a purchasing error.
Amazon Web Services recently published guidance that spells this out clearly: "Product information scattered across multiple systems, inconsistent data formats, and real-time inventory challenges create barriers that prevent AI agents from effectively representing retailers' offerings."
The stakes are high.
As Karen Webster, CEO of PYMNTS, wrote: "The challenge is no longer about being indexed but about being shortlist-eligible, which requires structured catalogs, clean metadata and semantic alignment."
Gartner forecasts that by 2030, at least 25% of all purchasing decisions will be made by machines. They've identified AI-driven buyers as "autonomous customers" making purchases based on rules, data, and optimization rather than emotional appeal.
Here's how SAP recommends retailers prepare: "Ensure product feeds are rich, structured, and machine-readable, complete with attributes, use-case-driven descriptions, real-time pricing, and accurate inventory. Clean, structured product data is now the foundation of intelligent discovery."
This represents a major operational shift for every business, particularly with the rise of ai powered purchases . The primary focus must move from optimizing for human eyes to optimizing for algorithms. This means investing heavily in data governance and ensuring every product has rich, structured attributes. In this new world, your data is no longer just an asset; it is the primary interface for commerce.
How Brands Should Prepare: A Strategic Framework
Based on my research and hands-on experience building AI tools, here's what I recommend for marketing leaders navigating this transition:
- Audit Your Product Data Immediately. Conduct a comprehensive evaluation of your product information completeness. Use data quality scoring for all products. According to McKinsey's research on agentic commerce, this requires "embedding semantic and behavioral metadata into product catalogs."
- Monitor Your AI Visibility. Start tracking how your brand appears in AI shopping assistants like ChatGPT, Perplexity, and Google's AI Mode. This is why I built tools like Prompt Runner—to help brands understand their visibility when AI agents are making recommendations. The brands winning tomorrow are those measuring this today.
- Prepare for Liability Questions. Work with your legal and finance teams to establish clear policies for AI-mediated transactions. Define who the "purchaser" is in agent-mediated transactions. This determination drives policies touching chargebacks, returns, warranties, and liability disputes.
- Optimize for Post-Purchase Experience. With discovery being augmented by AI agents, humans will give their ongoing loyalty based on post-purchase experiences. SAP's research found that 72% of consumers report they will only stay loyal to brands that consistently meet their needs in the moment.
- Build API Infrastructure. To survive, retailers must become "agent-preferred"—the brands algorithms favor when completing purchases. That means building open APIs for inventory visibility, maintaining transparent pricing, and ensuring reliable fulfillment.
The Consumer Perspective: What Shoppers Actually Think
Despite the technical complexity, consumer adoption is accelerating rapidly. Adobe's survey of 5,000 U.S. consumers found that 38% have already used generative AI for online shopping, with 52% planning to do so this year. Among those who have used AI for shopping, 92% said it enhanced their experience.
Even more telling: 87% said they are more likely to use AI for larger or more complex purchases. This suggests AI isn't just for routine purchases—it's becoming the go-to for high-consideration decisions where research really matters.
Riskified's global study of 5,400 consumers found that 73% are already using AI somewhere in their shopping journey. Trust in AI (36%) is rapidly approaching parity with trust in in-store associates (38%). Only 25% prefer to shop online without AI assistance.
The concerns are real but manageable: payment security (32%), privacy (26%), potential mistakes (18%), and loss of control (17%). What's notable is that the benefits—convenience, speed, better deals—are outweighing the concerns for the majority of shoppers.
The Window for Action Is Open. But Narrowing.
Agentic commerce is far more than a new technology or a more convenient way to shop. It represents a fundamental reshaping of markets, a redefinition of customer behavior, and a new frontier of legal and financial liability. It challenges the very concept of brand loyalty and forces businesses to prioritize machine-readable data over human-centric design.
While the transition is complex, its arrival is not a distant fantasy. Major players are deploying production-ready systems now: OpenAI's "Instant Checkout" is powered by the Agentic Commerce Protocol (ACP), Google has established the security foundation with its Agent Payments Protocol (AP2), and PayPal is enabling millions of merchants with its "Agent Ready" services.
Deloitte projects that 25% of enterprises using generative AI will deploy autonomous AI agents in 2025, doubling to 50% by 2027. Gartner predicts that at least 15% of day-to-day work decisions will be made autonomously through agentic AI by 2028.
As AI agents learn our preferences, anticipate our needs, and begin to manage our daily purchases, we must ask ourselves a critical question: Are we prepared for a world where an algorithm knows what we want better than we do?
For brands and retailers, the answer needs to be yes. The window for early-mover advantage remains open, but it won't last indefinitely. The foundational protocols are being established, the consumer adoption is accelerating, and the first wave of agent-mediated purchases is already here.
Those who prepare now, by cleaning their data, building their APIs, and understanding their AI visibility, will thrive. Those who wait risk becoming invisible to an entire category of purchasers.