KEY HIGHLIGHTS
- AI-driven U.S. e-commerce traffic surged 758% year-over-year between November and December 2025, according to Adobe.
- Roughly 81% of retail executives believe generative AI will weaken brand loyalty by 2027, per Deloitte’s 2026 Retail Industry Global Outlook.
- Bain & Company research shows consumers currently trust retailers’ on-site agents 3x more than third-party agents — but that trust window is closing fast.
- About 50% of retail executives anticipate the collapse of the current multi-step customer journey by 2027 as AI compresses it into a single interaction.
- McKinsey estimates agentic commerce will generate $3–5 trillion in global economic value by 2030.
- The winning strategy is likely a blend: participating in third-party AI platforms while building differentiated, proprietary agentic capabilities on owned channels.
Retail has survived a lot of disruptions — the rise of big box stores, the e-commerce revolution, the mobile shift. But what is happening right now with agentic AI may be the most structurally disruptive development the industry has ever faced. And this time, the disruption isn’t just about how people shop. It’s about who, or what, retailers will be selling to.
That’s not hyperbole. It’s where the trajectory leads, and the smartest retailers are beginning to reckon with it.
The Agent Is the New Customer
The numbers tell a striking story about how quickly AI is reshaping shopping behavior. Adobe’s 2025 Holiday Shopping report found that AI-driven U.S. e-commerce traffic grew 758% year-over-year in November 2025 alone. On Cyber Monday, AI traffic to U.S. retail sites jumped 670%. These aren’t marginal metrics; they signal a behavioral shift underway at scale.
The strategic question retailers are grappling with right now is deceptively simple: should they meet customers where they are, which is increasingly, inside external AI platforms like ChatGPT, Google Gemini, and Microsoft Copilot, or should they invest in building proprietary agentic experiences that keep shoppers on their own turf?
In 2026, many are trying to do both, and the risks of each path are becoming clearer by the day. Etsy, Target, and Walmart have joined forces with Gemini and Copilot to make products available on those platforms, building on earlier moves to list merchandise on ChatGPT. Amazon and Walmart have also invested heavily in their own consumer-facing AI assistants — Amazon’s Rufus and Walmart’s Sparky — to own more of the discovery and purchase journey directly.
There is every reason to think this will shake up retail in a major way, not unlike the impact the internet had on spurring ecommerce and changing consumer discovery and buying habits.
The Data Problem Nobody Wants to Talk About
Here’s what gets lost in the conversation about meeting customers where they are and the importance of personalization: every interaction that happens inside an external AI platform is an interaction retailers don’t own. The browsing behavior, the comparison queries, the hesitation before purchase — all of that rich behavioral data that used to flow back to the retailer will no disappear into someone else’s platform.
That is a significant shift and one retailers need to be factoring into their strategies. If Gemini or Claude or ChatGPT “owns” the customer discovery process, can see the comparison/decision making process, and is completely left out of the delivery process, the relationship with the customer is, well, largely nonexistent. Even if the data is shared by AI platforms where these interactions are taking place, chances are good the context of the data around the discovery/buying/delivery process, if shared at all, will be incomplete, which means personalization is largely impossible.
Deloitte’s 2026 Retail Industry Global Outlookunderscores the stakes: 81% of retail executives believe generative AI will weaken brand loyalty by 2027. That’s an astonishing figure. It reflects a growing recognition that when an AI agent is making purchasing decisions on a consumer’s behalf, optimizing for price, availability, and ratings rather than brand affinity, traditional loyalty levers stop working. Deloitte’s research
Bain & Company research on agentic AI commerce adds another dimension to this challenge, which revolves around trust. Today, consumers trust retailers’ own on-site agents three times more than third-party agents. That trust advantage is real and meaningful. But it is also a closing window. Retailers that don’t build differentiated, capable on-site agentic experiences quickly will watch that advantage erode as consumers grow more comfortable with third-party AI platforms completing purchases on their behalf.
From Consumer Relationships to Agent-to-Agent Commerce
The longer-term scenario is even more complex. We are nearing an age where those AI agents don’t simply research on a customer’s behalf, those agents will ultimately begin negotiating with a retailer’s AI agent as part of the buying process. Now, you’ll have two autonomous systems, transacting without a human in the loop. For a frequent traveler, this could mean an AI agent learns preferences, books flights, arranges accommodations, and handles logistics entirely autonomously — the user never even opens the app. As a frequent traveler myself, I’ll admit that this makes me slightly nervous but also excited at the possibilities of handing over this time-consuming task.
Translate that to retail: a shopper’s AI agent, configured with style preferences, size, budget parameters, and brand loyalties, interacts directly with a retailer’s agent to complete a purchase. What does that mean for the role of merchandising, marketing, and brand experience? It means they need to be optimized for agent consumption, not just human appeal.
Interacting with the customer will be a thing of the past: interacting with their AI agent representative will quickly become a new reality. For brands, that means taking into consideration how information is presented online, how agents will interact with that information, and strategies that can be used to “persuade” AI agents that your goods are in some way a better option than the competition.
In short, it’s a whole new ballgame and a profound operational shift. Product metadata, pricing transparency, inventory accuracy, and structured data quality become competitive differentiators — because those are the signals an AI agent will use to evaluate and choose. Research from Airia found that enriching product data with semantic layers can improve AI model accuracy from 16% to 54%. That’s the kind of infrastructure investment that will separate winning retailers from those who get commoditized.
The Strategies Emerging from the Chaos
The good news is that retailers get the massive shift that’s underway and aren’t sitting still. They are pursuing several parallel strategies to maintain relevance and control in this environment, and the smartest players are hedging across multiple fronts simultaneously.
Some are doubling down on proprietary agentic capabilities on their own platforms. Home Depot’s AI companion, Magic Apron, is exclusively available on their website, giving it access to customer data and purchase history that third-party agents can’t replicate. That creates a genuine differentiation advantage and protects against the commoditization threat. Here’s a look at a reddit post that shows how it works (and the value it delivers). Without questions, there is an immediate benefit to be able to ask an AI agent your questions before purchasing, without having to wade through customer reviews and FAQs to get quick answers. I’ll note that doesn’t mean that customer reviews aren’t still incredibly valuable as part of the discovery/purchase consideration process, but this is some cool functionality that can easily provide an assist in the shopping process.
Others, like Walmart, are hedging by participating in external platforms while also investing in its own AI assistant, Sparky, to capture direct relationships. Amazon has taken the most aggressive stance — building its own Buy for Meagent that can shop other brands’ websites and simultaneously restricting external agents from accessing Amazon’s own storefront. That last point is a smart, if risky, power move and not at all surprising as it relates to battling the OG ecommerce giant.
Google’s announcement at NRF of its Universal Commerce Protocol (UCP), an open source standard designed to power the next gen of agentic commerce is an example what some have called “defensive innovation in the AI era.” UCP is an understandable attempt by the search giant to hold onto its long-established dominance as a search and discovery engine. The framing of the UCP announcement was cooperative, merchant-centric, and designed with ease-of-use top of mind. Google’s Sundar Pichai emphasized that retailers remain the merchant of record throughout the process, preserving their ability to personalize offers and shape the customer relationship. Google’s UCP is designed as an ecosystem strategy not the least of which reduces integration friction, integrates AP2’s dual mandates of intent and cart) and features OAuto 2.0 to link identities. This is important, as it makes every interaction auditable and makes trust in AI-driven commerce a bit easier to come by. Merchants will need to shift their focus from SEO-driven strategies to optimizing for agentic interactions, which will impact how product and promotional information and service capabilities are available to and understood by AI agents.
Whether the practice lives up to that promise will depend on what data actually flows, and in which direction.
I agree with Bain & Company’s analysis, which suggests the winning approach for most retailers won’t be a binary choice. Smart players will participate in third-party agentic platforms to maintain visibility and reach, but simultaneously invest in on-site agentic capabilities, protect data and checkout control, and build differentiated experiences that give consumers a reason to engage directly.
The Bottom Line: This Is an Early-Stage Problem With Long-Tail Consequences
About half of retail executives anticipate the collapse of the traditional multi-step shopping journey by 2027, replaced by a single AI-driven interaction. McKinsey estimates agentic commerce will generate $3–5 trillion in global value by 2030. IDC projects agentic AI will represent 10–15% of enterprise IT spending in 2026, growing to 26% of budgets by 2029.
We are genuinely in the early innings of this disruption, but early innings don’t mean safe innings. The choices retailers make right now about where to participate, where to invest, and how to protect their data and customer relationships will determine their competitive position for the next decade.
The parallel to ecommerce’s emergence in the 1990s is apt. Retailers that dismissed online channels as marginal fell behind. Retailers that built strong, differentiated digital capabilities alongside their physical presence were better positioned for long-term success. The same dynamic is playing out today, just at a significantly compressed timeline.
AI isn’t just changing how customers shop. It’s changing who shows up to shop in the first place. Retailers that recognize the agent as a new class of customer — one that needs to be understood, engaged, and won — will be the ones best positioned to thrive in what comes next.
Read more of my coverage:
Agentic Commerce Is Coming — and AWS Is Building the Payment Rails to Get It There
This article was originally published on LinkedIn.