AI in retail has crossed a line that most shoppers have not noticed yet: the software recommending your next purchase is starting to complete the purchase itself. At the National Retail Federation’s Big Show in January 2026, Google chief executive Sundar Pichai stood alongside Walmart’s incoming chief executive John Furner to announce the Universal Commerce Protocol, an open standard built with Shopify, Target, Etsy, Wayfair and Walmart that lets an AI agent discover a product, apply a loyalty offer and complete checkout without the shopper ever leaving a chat window.
In his own published remarks from the event, Pichai said Google’s systems processed 8.3 trillion shopping-related tokens in December 2024 and more than 90 trillion a year later, an elevenfold jump he called proof of how fast the discovery-to-purchase journey is changing.
That shift matters now because it is no longer a pilot programme confined to a handful of tech-forward retailers. Agentic commerce, AI systems that can browse, compare and buy on a customer’s behalf rather than simply suggest what to buy, is moving from conference keynote to checkout page inside a single calendar year. Understanding what AI in retail actually changes for shoppers, store staff and the retailers racing to keep up is now a practical question rather than a speculative one.
What AI in Retail Actually Looks Like at Checkout
The clearest evidence of how AI in retail is being deployed sits inside Google’s own Universal Commerce Protocol, known as UCP. Once a retailer adopts it, a shopper asking Gemini or Google’s AI Mode about a product can be shown a personalised offer, such as a new-member discount or a loyalty enrolment prompt, and can finish the purchase through Google Pay without ever opening the retailer’s own website or app.
The retailer remains the merchant of record throughout, which means it still owns the customer relationship and the transaction data, even though the AI system handled the entire discovery and decision journey.
Walmart moved quickly to plug into that infrastructure. The retailer paired its Sam’s Club and Walmart assortments directly into the Gemini app, and, as the National Retail Federation’s own coverage of the announcement describes, Furner joined Pichai on stage specifically to discuss what agent-driven commerce means for a retailer of Walmart’s size. Pichai’s own remarks named The Home Depot, McDonald’s and Kroger as early partners already using a related Google tool, Gemini Enterprise for Customer Experience, to fold shopping assistance, support and merchandising into a single agent-driven interface rather than three separate systems.
LiveAIWire’s reporting on how a recommendation algorithm now shapes what Hollywood makes found a similar dynamic already at work in entertainment, where an AI system quietly restructures what gets surfaced to a person long before that person consciously decides anything. Retail is now running the identical experiment against your shopping cart instead of your watchlist, and the stakes are more direct: a buried film recommendation costs you an evening, but a buried product recommendation can cost a retailer the sale outright.
Why Retailers Are Racing to Adopt It
The commercial logic is straightforward even where the technology is new. Industry analysis from PYMNTS found that both Amazon and Walmart effectively swapped scripts during the opening weeks of 2026, with the historically closed platform, Amazon, embracing a new agentic standard while the historically open one, Walmart, pushed deeper into owning the full conversational purchase journey inside its own ecosystem. Neither company wants to be the retailer whose products an AI agent cannot discover, compare or check out with, because the alternative is invisibility inside a shopping journey that no longer starts on a retailer’s own homepage.
That invisibility risk is why more than twenty launch partners, including Visa, Mastercard, Stripe, Best Buy and The Home Depot, signed onto UCP within months of its announcement. LiveAIWire’s analysis of the difference between generative and predictive AI found that most enterprise AI spending still fails to show up as measurable earnings impact, which is exactly the trap agentic commerce is built to avoid. A shopping agent that completes an actual transaction, rather than simply drafting a paragraph or summarising a document, produces a number a retail finance team can point to directly: a completed sale, not a productivity anecdote.
How Big This Shift Already Is
The numbers behind agentic commerce are moving faster than most retail strategy documents can keep up with. Analysts covering the sector project that agentic AI could represent up to 15 percent of retail IT spending in 2026 alone, climbing toward roughly a quarter of all retail technology budgets by the end of the decade. Coresight Research has described checkout-enabled AI agents as only the opening chapter of what it calls an agentic universe, where a single digital assistant increasingly manages discovery, comparison, purchase and post-purchase support in one continuous conversation rather than four separate visits to four separate pages.
Retail media networks are being pulled into the same shift. Amazon Ads and Walmart Connect already account for the overwhelming majority of incremental retail media spending, and Walmart Connect is projected to be the only major network still gaining share through 2027, according to industry forecasting cited by EMARKETER.
That dynamic reinforces the same pattern already visible in the UCP launch: retailers with the scale to build or plug into an agentic layer are pulling further ahead of smaller competitors still running a conventional website and a static loyalty app, and the gap between the two groups is compounding every quarter rather than narrowing.
The Personalisation Layer Behind Every Recommendation
Underneath the agentic checkout layer sits a much older and more mature technology: predictive AI models that have been forecasting demand, managing inventory and scoring fraud risk in retail for the better part of two decades. Those systems now feed the newer generative layer that drafts personalised offers and holds a conversation with the shopper. LiveAIWire’s coverage of how organisations choose the right AI model for a given task found that matching the tool to the job, rather than defaulting to the most capable system available, is the highest-leverage decision most AI programmes make all year.
Retail’s agentic commerce boom is a direct application of that principle at industry scale: a predictive engine decides what a customer is likely to want, and a generative layer turns that prediction into a natural conversation and a completed purchase.
The same pairing of prediction and generation that now drives your Netflix queue or your Spotify playlist is what decides which loyalty offer appears when you ask Gemini about a new pair of trainers. The difference in retail is that the loop closes with money changing hands in the same conversation, rather than with another hour of your attention.
What This Means for the People Who Work in Stores
Agentic commerce does not only change what happens on a shopper’s phone. It changes what retail staff are asked to do, and who is still needed to do it. LiveAIWire’s reporting on AI job displacement and augmentation across the wider economy found that clerical and routine transactional roles carry the highest measured exposure to generative AI tools, according to a March 2026 International Labour Organization index, while roles that depend on physical presence, judgement in unpredictable situations and interpersonal trust remain largely insulated.
Retail sits uncomfortably across both categories. Routine online order processing and basic customer service chat are squarely in the exposed group. In-store roles that require handling an awkward return, resolving a delivery dispute a chatbot cannot untangle, or simply being a trusted human presence on a shop floor look far more durable.
What is shifting inside stores that keep their staff is the nature of the job itself. Warehouse and fulfilment workers increasingly work alongside AI systems that optimise picking routes and predict which items need restocking before a shelf actually empties, augmenting the job rather than eliminating it outright. The workers most exposed are not necessarily the ones on the shop floor. They are the back-office analysts whose forecasting and reporting tasks a predictive model can now perform directly, a pattern consistent with what LiveAIWire’s wider labour market coverage has found across other white-collar functions facing the same transition.
The Risk AI in Retail Has Not Solved Yet
None of this is friction-free. A pricing algorithm optimised purely for margin can quietly treat two customers with identical needs very differently, based on correlated signals about income, location or shopping history that no one directly entered into the system. LiveAIWire’s coverage of how AI already reshapes insurance pricing found the same structural tension in a different sector: a model built purely to maximise predictive accuracy can misread anyone who does not fit the statistical average, and the people most often misread are already on the margins of the market.
Retail’s agentic pricing engines face an identical test, quietly deciding which customers see a discount and which see a higher price for the same product, based on data patterns the shopper never consented to in any explicit sense.
There is also a simpler, more immediate concern: a purchase completed entirely inside a conversational interface removes the pause that a shopping cart page used to provide. When comparing, deciding and paying all happen inside the same exchange with an AI agent, the friction that once gave a shopper a moment to reconsider an impulse purchase disappears along with the inconvenience it was designed to prevent. Retailers building agentic checkout flows have an incentive to treat that removed friction as a feature. Regulators and consumer advocates are increasingly treating it as a risk that current disclosure rules were not written to cover.
What This Means for You as a Shopper
For most people the practical change is already here rather than hypothetical. If you have asked Gemini, ChatGPT or a retailer’s own chatbot a shopping question in the past few months, there is a reasonable chance an agentic system, not a static webpage, shaped what you were shown and how easy it was to buy it. Reading a platform’s data use disclosures before linking a loyalty account to a conversational assistant is a more useful habit now than it was two years ago, precisely because the systems making purchase suggestions are increasingly the same systems completing the transaction.
Comparison shopping across more than one AI surface, rather than trusting a single assistant’s first suggestion, remains the most reliable defence against a system quietly optimised for a retailer’s margin rather than a shopper’s actual interest. The retailers investing hardest in agentic commerce right now, Walmart and Amazon foremost among them, are the ones defining what the next decade of retail competition looks like. Whether that competition produces a genuinely better shopping experience or simply a faster, less visible one is a question that current adoption data cannot yet answer, and the next eighteen months of UCP rollouts, checkout data and regulatory scrutiny will begin to settle it either way.
About the Author
Stuart Kerr is Technology Correspondent at LiveAIWire, covering artificial intelligence, emerging technology, and their impact on business, society, and everyday life. LiveAIWire publishes original AI journalism every weekday at liveaiwire.com.