Vernon Yai is a seasoned authority in data governance and protection, specifically navigating the complex intersection of retail operations and emerging AI technologies. With years of experience helping brands secure their most sensitive information while optimizing it for utility, he understands that the true barrier to AI adoption isn’t the software—it’s the integrity of the data underneath. His focus remains on bridging the gap between “data debt” and high-performing autonomous commerce, ensuring that organizations can move beyond costly pilots into sustained revenue growth.
In this discussion, we explore the significant performance discrepancies found in early automated checkout pilots and why leading retailers are pivoting from transaction-led AI back to discovery-focused agents. We delve into the technical shifts required to move from linear session-based analytics to persistent customer threads, the impact of fragmented data on brand trust, and how real-time context serves as a brand’s ultimate competitive moat in a world of commoditized AI models.
Automated checkout pilots often show conversion rates significantly lower than traditional web storefronts. Why does this performance gap occur, and what specific technical friction points must be addressed before an AI agent can rival a standard e-commerce site?
The reality is quite stark, as seen in major trials where automated checkout conversions were roughly 3X worse than traditional sites. This gap exists because most current AI agents act as little more than expensive search bars, lacking the deep flexibility required to handle the nuances of a real purchase. To rival a standard site, we must first address the “contextual friction” where the agent doesn’t know who the shopper is across different devices. Technically, this requires moving away from overnight batch updates to real-time synchronization so the agent never promises a product that went out of stock ten minutes ago. Without a unified data layer that connects identity, inventory, and promotions, the experience feels “unsatisfying” and clunky compared to the streamlined flow of a modern web store.
Many AI platforms are pivoting from handling direct transactions back to focusing on product discovery. How does this shift change the data requirements for retail leaders, and what are the trade-offs of letting an external agent manage the discovery phase versus the final purchase?
When the focus shifts to discovery, the data requirement moves from simple transaction security to rich, persistent preference signals. It’s no longer just about processing a credit card; it’s about the agent understanding a customer’s style, past returns, and even their loyalty tier in real time. The trade-off is that while an external agent can offer massive reach and high-funnel engagement, the brand risks losing control over the “connective tissue” of the customer journey. If the external agent is discovery-focused but doesn’t have access to the merchant’s real-time inventory, it might lead a customer to a “dead end” product, which chips away at the trust the brand has worked to build. To succeed, retailers must provide these external agents with a trusted view of their product data so the discovery feels considered rather than generic.
Agentic commerce is projected to capture a massive share of the market by 2030, yet many shoppers move across multiple touchpoints before buying. How can businesses unify these fragmented sessions into a single customer thread, and what specific metrics indicate a successful integration?
With the market projected to hit between $300 and $500 billion by 2030, we have to stop treating each interaction as a fresh start. Currently, a shopper might research on a phone during a commute and then switch to a laptop, which legacy systems view as two separate people. Unifying these sessions requires an architectural shift where the “customer” is the primary thread, not the “session,” allowing the agent to recall a cart from three days ago on an entirely different device. A key metric for success is “context intelligence,” or the frequency with which an agent can provide a recommendation that acknowledges a cross-channel action, such as a recent in-store return. When an agent can proactively mention a promotion that matches a user’s loyalty status during a discovery session, you know the integration is functioning at a high level.
Most retail systems assume a linear path from “add to cart” to checkout, yet real-world behavior is disjointed. What structural changes are needed to move away from session-based analytics, and how does this transition impact real-time inventory visibility?
The biggest structural change is moving identity resolution from a back-office administrative task to a live, customer-facing capability. Our systems were built for how we wished customers shopped—arriving, browsing, and buying in one go—but real life is messy and non-linear. This shift demands that inventory visibility be absolute and instantaneous; you cannot have a siloed supply chain where a bundle ships in two pieces from different nodes because the data was stale. For example, if a customer asks an AI assistant about a product they saw in a physical store, the system must immediately reconcile that store’s stock with the online availability. If there is a delay in that synchronization, the agent will inevitably make broken promises regarding delivery windows or stock levels.
Roughly half of technology leaders report that their data stacks are not ready for AI agents. What specific “data debt” typically causes these systems to fail, and what initial steps should be taken to resolve customer identity across web, app, and physical store environments?
That 50% figure is a direct result of decades of siloed growth where the mobile app, the website, and the brick-and-mortar stores were treated as different businesses. This “data debt” manifests as fragmented IDs, where a single person has three different profiles depending on where they shop, making it impossible for an AI to offer a coherent experience. The first step to resolving this is a rigorous unification of customer and product data into a single, trusted source that the AI can draw from in real time. We have to move past the era of batch updates and embrace a data layer that acts as the “connective tissue” between all touchpoints. Once you have a unified identity, the AI can finally stop recommending items the customer returned last week, which is one of the most common points of failure in today’s pilots.
As AI models become commodity tools, a brand’s unique context becomes its primary advantage. How does real-time synchronization of order history and loyalty status differentiate an AI experience, and how can a firm ensure this data remains consistent during high-traffic periods?
Since every retailer will eventually have access to the same powerful foundation models, the model itself is no longer the competitive edge; the quality of your context is. When an AI agent can instantly access a shopper’s loyalty status and order history, it transforms from a generic chatbot into a brand-differentiated personal assistant. Ensuring consistency during high-traffic periods requires a robust data infrastructure that doesn’t buckle or revert to “stale” cached data when things get busy. If an agent commits to a specific price or delivery window during a Black Friday rush, that data must be synchronized across every fulfillment node instantly. If the context is wrong, even the most advanced AI model in the world will provide a recommendation that feels generic and uninspired.
What is your forecast for agentic commerce?
I forecast that we are entering a period of “The Great Rationalization,” where the hype of AI agents will be met by the cold reality of data maturity. By 2030, agentic commerce will indeed represent up to 25% of the market, but those gains will be concentrated among the few retailers who prioritized their data layer over the model layer. We will see a shift where “context intelligence” becomes the primary KPI for CIOs, as the ability to provide a unified, real-time picture of the business becomes the only way to make AI profitable. The retailers who fail to resolve their data debt will find themselves trapped in perpetual, costly pilots, while those with unified data will turn their agents into the most effective sales tool the industry has ever seen.
