I met “Mike” at a trade show: a bleach‑blond hologram in a pink suit, suspended in a clear tube, answering audience questions on a three‑second delay. He was charmingly awkward — built more to lure browsers than to hold a real conversation — and he summed up the mood on the National Retail Federation floor. Everywhere you looked, companies were pitching ways to drag more of the shopping journey into artificial intelligence: discovery, recommendations, checkout, even the checkout that happens inside an assistant rather than on a retailer’s own site.
A new infrastructure for buying
The biggest change at NRF and CES wasn’t another shiny gadget. It was infrastructure: standards and agent templates designed to let third‑party AIs act on behalf of shoppers. Google unveiled an open protocol to let agents and retailers speak the same commerce language, and rolled out tools that let brands place virtual sales associates inside its AI experiences. Microsoft answered with its own checkout and brand‑agent tools for Copilot, integrating partners like Stripe and Shopify so a purchase can be completed inside an assistant.
Those moves are practical. They shrink friction: you can describe what you want conversationally and let an agent find, compare and buy it. But they also rearrange power. As retailers let their catalogs and inventory be surfaced inside Google or Copilot, the platforms sit between shopper and store — and they control the logs, the recommendation signals and, crucially, the data trails that explain why a purchase happened. For context on how fast these platform plays are spreading, see recent developments in agentic booking and Google’s AI Mode and Google’s broader Workspace integrations that pull Gemini into Gmail and Drive workflows here.
Who owns the customer?
That question — who controls the discovery and the data — was the undercurrent of conversations on the show floor and onstage. Retail leaders at NRF were upbeat about the value AI can add: better personalization, faster inventory forecasting, even smarter same‑day delivery orchestration. But in quieter moments, execs warned about disintermediation. If an assistant recommends a product and takes the payment, what does the retailer keep? What analytics, what context, what follow‑up rights?
Some vendors promise to help brands monitor their exposure and placement inside chat‑first search; others offer tools to reverse‑engineer when and why an AI surfaces certain items. All of it smells like a new wave of optimization — except this time the battleground is not search result ranking but how models answer you, and who gets credited when they do.
The surveillance layer — online, and increasingly in stores
AI’s reach is not limited to where you type. Companies showed in‑store systems that pair cameras, attention metrics and identity tokens to track how long someone glanced at an ad, whether they picked up a product and how that interaction maps to later purchases. Vendors say video is deleted after metadata is captured; civil liberties groups and wary shoppers note that ‘metadata’ still paints an intimate portrait of behavior.
Meanwhile, personalization is getting more ambitious — and creepier. Pizza chains demoed assistants that can autofill your usual order, suggest items for a crowd and even estimate how many people will eat based on a photo you upload. That kind of convenience is useful. It also makes it easier for platforms and brands to stitch together identities across channels.
The human cost, and the human counterpoint
Not every exhibitor was screaming about automation. A small packaging company quietly showcased tactile designs — reusable dustbags, insulated totes — and made a persuasive case: some parts of commerce are still about touch, not algorithms. At a time when “agentic” tools promise scale, that physical intimacy can be a differentiator.
Retail executives’ public comments reflected a similar tension. Some urged rapid experimentation — “fail fast,” one C‑suite leader advised — while others counseled caution, saying AI should augment associates rather than replace their knowledge. Several leaders argued the smartest use of AI is to free humans for higher‑value work: deeper product obsession, better in‑store service, creative merchandising.
Ad tech, agencies and the automation of marketing
It’s not just stores and checkout. Agencies and ad tech firms are embedding agentic AI into media planning and programmatic buying: templates that build campaign plans, operating systems that bid in real time according to high‑level goals. That will change how budgets are spent and how creative is iterated — faster, but potentially less human in the loop.
Brands will have to decide what they hand over to automation and what they keep for people who understand nuance, context and culture.
A practical guide for retailers (and shoppers)
If you work in retail: prioritize the data contracts now. Decide what you must retain — customer identifiers, conversion paths, post‑purchase signals — before you let third‑party agents handle transactions. Test with guardrails. And don’t lose the tactile hooks that build loyalty offline.
If you shop: expect more convenience, more personalization, and more things you didn’t know you wanted. Also expect to make more uncomfortable tradeoffs about privacy and control. Read privacy policies when you can, and ask brands where your data lives.
The next wave of retail will be fast, conversational, and increasingly invisible — purchases sealed with a sentence, not a cart page. That convenience is seductive. But in the rush to automate, the real question is not whether AI can sell us more, but who gets to keep the map of how we were sold.