This article was based on the interview with Tealium’s Zack Wenthe on the hidden challenge of AI-driven commerce by Greg Kihlström, AI and MarTech keynote speaker for The Agile Brand with Greg Kihlström podcast. Listen to the original episode here:
The industry is abuzz with the potential of agentic commerce, the seamless integration of shopping and purchasing capabilities directly within AI chat interfaces. The allure is undeniable: meeting customers where they are, offering frictionless discovery, and enabling instant checkouts without ever forcing a user to leave their chat window. For marketing leaders who have spent the last decade optimizing every click and pixel of the on-site customer journey, this represents a monumental shift. It’s a chance to be at the bleeding edge, to redefine convenience, and to tap into a rapidly growing channel where high-intent customers are actively seeking solutions.
However, beneath the surface of this exciting new frontier lies a significant challenge that should give every data-driven leader pause. As we race to enable these AI-powered shopping experiences, we risk creating a new and formidable data black box. When the crucial stages of research, comparison, and consideration happen inside a third-party AI environment, the rich behavioral insights we’ve come to rely on can become disconnected from the overall customer journey. This isn’t merely a new channel to track; it’s a fundamental shift in visibility that threatens to make our understanding of the customer path-to-purchase less clear, not more. The strategic question is no longer just if we should participate, but how we can participate without sacrificing the intelligence that fuels our entire marketing engine.
The New Walled Garden
The core of the issue is a loss of control. For years, marketers have worked to centralize customer journey data, meticulously tracking how a user moves from an ad to a landing page, browses product categories, and proceeds through checkout. Agentic commerce fundamentally disrupts this model by moving the most valuable parts of that journey off-platform and into an environment owned by a third party.
“Your whole, you know, kind of checkout journey experience is shifting out of your environment in your control in a lot of ways to a walled garden, to a third party who is handling the discovery, is passing on information to you, but obviously you don’t know how that consumer browsed, what the decision criteria was, what they were comparing it against potentially. And so you miss out on a lot of that shopping experience.”
Wenthe’s point here is critical. This isn’t just another referral source like social media or search. In those cases, the discovery may happen elsewhere, but the substantive interaction and data collection still occur on owned properties. In the world of agentic commerce, the AI is not just a referrer; it’s an active participant and gatekeeper. It interprets the user’s needs, curates a selection of products (based on its own opaque algorithms), and facilitates the decision. By the time a customer either clicks through to a product page or completes a transaction via an API, the “why” behind their choice is largely lost. For leaders who rely on this behavioral data to inform product development, merchandising, and personalization strategies, this is a significant blind spot.
A Shift in the Definition of a Touchpoint
This new paradigm forces us to redefine what we even consider a customer touchpoint. If a customer has a detailed conversation with an AI about your product category, compares your offering to three competitors, and discusses price, features, and shipping—all without ever interacting with your brand directly—is that a touchpoint? The interaction is certainly happening, but your ability to observe or influence it is severely limited. Wenthe offers a fitting, if slightly unsettling, analogy.
“It’s almost as if say a retail store, it’s like a guy standing outside your front door telling you about couches before you walk in the door. And you’re like, I don’t have a lot of control over what he’s saying. I’m just hoping they eventually walk through my door. And that’s going to be a shift for a lot of organizations.”
This metaphor perfectly captures the anxiety many brands feel. The “guy outside the door” is the LLM, representing your brand based on the data it has scraped and the prompts it receives. You have no direct control over how your products are described, what competitors they are positioned against, or whether the information presented is entirely accurate or favorable. This elevates the importance of foundational, structured data and content marketing. The information on your site must be so clear, well-organized, and accessible that the AI can interpret it correctly. It’s a move from actively managing the customer experience to passively influencing a third-party intermediary, a shift that requires a different set of skills and a renewed focus on technical SEO and content strategy.
Surviving on a New Data Diet
Given these limitations, what data can brands actually collect, and what architectural changes are needed to capture it? Wenthe outlines two primary scenarios, each with its own data profile. The first is a more traditional referral, where a user clicks a link in a chat and lands on your site. The data here is sparse but valuable.
“In that scenario, what’s happening is it’s almost high intent but low directionality data that’s coming in because somebody’s landing cold on your website and then almost immediately going into or hopefully immediately going into some sort of buying pattern. You don’t know really what led them there.”
This “high intent, low directionality” traffic requires a different approach. These visitors bypass the traditional top-of-funnel content. They arrive ready to transact, meaning the window to understand and influence them is small. This is where a robust first- and zero-party data strategy becomes non-negotiable. Post-purchase surveys (“How did you find us?”), on-site polls, and progressive profiling become essential tools to backfill the context that is missing from the referral data. A Customer Data Platform (CDP) becomes critical not just for unifying touchpoints, but for quickly associating this anonymous, high-intent behavior with a known profile and triggering real-time experiences to gather more information.
The second scenario is even more data-deficient: direct checkout via a commerce protocol where the transaction happens entirely within the chat interface. Here, the data is purely transactional.
“You just get basically fulfillment information. You get the customer name, you get the information, you get the product, and now you’re missing out on a whole lot of experience… you don’t get to inject coupons or promotions to maybe influence basket and purchase size, you don’t have loyalty information.”
This is the trade-off for ultimate convenience. Brands gain a sale but lose all the adjacent data and optimization levers—upsells, cross-sells, loyalty point integration, and basket-building promotions. As Wenthe notes, this is a key reason some major retailers are proceeding with caution. The lower average order value from single-item, AI-driven checkouts may not justify the investment, especially if it cannibalizes higher-value traffic from other channels.
The Unexpected Resurgence of Brand
In an environment where performance marketing data is becoming scarcer and less reliable, the strategic importance of brand marketing sees a powerful resurgence. When you can no longer meticulously optimize every step of the digital journey, you must instead focus on influencing the very first step: the user’s prompt. The way a consumer frames their request to an AI can dramatically alter the results.
“The way a consumer prompts the LLM is going to have a dramatic impact on the results you get, the products that get served. So if I start with a search for, you know, ‘hey, I’m looking for a couch kind of like X, Y, and Z,’ that’s going to be a very, very different kind of like criteria and… how you’re getting compared against, ‘hey, I’m just looking for a leather couch.’”
This insight is a call to action for every CMO. If a consumer begins their search by referencing your brand by name, or by using terminology and attributes that you have successfully associated with your brand, you have already won a significant battle. This places a premium on top-of-mind awareness and building strong brand equity outside of direct-response channels. Furthermore, Wenthe points to the concept of “memory” within LLMs, where the AI learns a user’s preferences and brand affinities over time. Getting customers to talk about your brand to their AI assistants could become a powerful, albeit indirect, marketing objective. The goal is to be so ingrained in your customer’s mindset that they carry your brand with them into these new walled gardens.
The move toward agentic commerce is not a trend to be ignored, but it must be approached with a clear understanding of the new data landscape. It introduces a level of abstraction between brands and consumers that we haven’t seen since the rise of third-party retail. The journey ahead will be one of careful experimentation, where success is measured not just by transaction volume from a new channel, but by our ability to mitigate the accompanying data loss. It requires a renewed commitment to collecting first-party data, an obsessive focus on structured content, and a deep investment in brand building.
Ultimately, the leaders who thrive in this new era will be those who embrace a dual strategy. They will build the technical agility to integrate with these new platforms and capture what limited data is available. At the same time, they will double down on the timeless principles of marketing: creating a brand so resonant that customers advocate for it, even when they are just talking to a machine. The future of the customer journey may be happening in a black box, but the power of a strong brand can still be the light that guides them to your door.
