I lead a marketing department in B2B. That's the day job: buying groups, pipeline, long sales cycles, getting leadership to see what marketing is worth. But I keep my finger on the pulse of all of marketing, consumer and retail included, because the most useful ideas rarely stay in the lane they started in.
So when I read Reuters' analysis this week on whether AI shopping agents could threaten Amazon and Walmart's advertising business [1], I immediately noticed something most of the coverage was missing.
The story itself is a fight over who gets to shop. Amazon is keeping outside AI agents off its store. Walmart is building a door for them. Google is putting more ads inside its AI answers. And most of the commentary has boiled it down to one question: is retail media doomed, or is it fine?
Both camps are looking at the wrong spot. Agents don't remove the buying decision. They move it. For twenty years the deciding moment happened at the shelf, physical or digital, which is exactly where retail media sells its space. When a customer hands the shopping to an agent, the deciding moment moves earlier, into a single sentence: the brief they give the agent. Whoever wins that sentence wins the sale, and very little of today's retail media budget is aimed at it.
Three things, all in the same few weeks.
The two biggest retailers chose opposite strategies. Amazon banned Meta's shopping agent, Muse, from its site in September, saying it had accessed the store without authorization, and it is still fighting in court over Perplexity's Comet agent after an appeals court overturned an earlier ban [1]. Walmart went the other way. On October 21 it rolled out infrastructure for third-party agents: when a shopper uses one, Walmart creates a profile for that agent linked to the shopper's account, and the shopper can approve purchases or set spending limits [1].
The money started to look less certain. Amazon's advertising revenue grew 11% in the third quarter, its slowest pace in nearly four years according to LSEG data [1]. To be fair, Andy Jassy said outside agents have had no meaningful effect on Amazon's ads yet, and a Morningstar analyst called this “an evolution rather than an existential threat” [1]. But Morgan Stanley researchers estimate agents could handle 10% to 20% of US e-commerce spending by 2030, and US retail media is a roughly $74.4 billion market, 88% of it held by Amazon and Walmart [1]. Nobody with that much at stake waits for the effect to become meaningful.
The platforms started rebuilding the shelf inside the conversation. Walmart is testing ads in Sparky, its shopping chatbot, and Amazon has put ads into Alexa for Shopping [1]. On October 28 Google expanded ads in AI Mode to ten more countries and launched “multiturn” ads that appear on follow-up questions in the US. Existing Search, Shopping and Performance Max campaigns are eligible automatically, with no separate opt-out [2]. Some users have since reported seeing four sponsored ads under a single AI Mode answer, which looks like a test [3].
Read those three together and the pattern is clear. Every platform is racing to own the moment where a person's intent turns into a purchase, because that moment is moving.
To the brief. The sentence a customer says or types to an agent is the last moment a human is fully in the loop. After that, a machine is comparing options on their behalf.
Not every brief works the same way, and the difference matters for where you spend. There are three kinds.
Interactive · The three briefs
Tell the agent what to buy.
Pick the sentence your customer is most likely to say. See what decides the sale.
Pick a brief to see what decides the sale.
Notice what is missing from all three: the moment of hesitation at the shelf. That moment, where a shopper compares two products and a well-placed sponsored listing tips the choice, is the moment retail media was built to sell. A named brief skips it entirely. A specced brief hands it to a machine that weighs data, not creative. A delegated brief replaces it with habit.
That is why the platforms are putting ads inside the agents themselves. They are trying to put the shelf back into a conversation that no longer needs one.
Stop treating this as a retail media question and treat it as a budget architecture question. Each brief has its own lever, and most budgets are only funding the shelf in between.
When a customer names you, you have already won, and the agent just does the paperwork. This makes brand salience more valuable in an agent world, not less. The spend that gets cut first in a tight year, the work that makes your name the one people say out loud, is the only defense against an agent quietly picking a cheaper substitute.
The test is simple. In your category, when someone describes the job to be done, does your name come out of their mouth? If not, no amount of shelf placement will get you into a brief that never mentions you.
When no brand is named, the agent compares, and it compares what it can read. Procter & Gamble, Nestlé and L'Oréal have all said they are working to make product information digestible for AI agents [1]. That is the tell. The biggest advertisers in the world are now treating product data as a marketing asset.
For most companies this means marketing has to care about things it used to hand off to e-commerce or operations: complete specs, accurate pricing and stock, clear delivery and return terms, honest reviews. An agent can't be charmed by a hero image. It can be disqualified by a missing size chart.
“Restock the usual” is decided by whoever holds the purchase history. That is the real stake in the Amazon and Walmart split. Whoever owns the shopper's account owns the default, and the default is the most valuable position in agentic commerce.
For brands, the move is to earn the reorder before the agent arrives: subscriptions, loyalty, replenishment programs and a product good enough that nobody wants to switch. Every repeat customer you hold today is a delegated brief you win tomorrow.
This one is specific to 2027 planning. One retail media executive told Reuters he expects Amazon and Walmart to start signing their largest advertisers to multiyear ad deals to secure that revenue [1]. If that offer lands on your desk, be careful about locking in today's shelf prices for years while the shelf is moving.
If you sign, negotiate for the world that is coming. Ask for reporting that separates agent-initiated orders from human ones. Ask what data you get back when an agent buys your product. And build in a way to re-price or exit if agent-driven sales become a meaningful share of your category. A retailer confident that its media will hold its value should not mind any of those terms.
This is where it comes back to my world, and why I think B2B teams should be watching consumer marketing closely right now. B2B buyers are already using AI assistants to research vendors and draft shortlists. The same three briefs exist: “get us a proposal from the firm we heard speak last month” is named, “find three vendors who meet these security requirements” is specced, and “renew what we have” is delegated.
The specced brief is where most B2B companies are weakest. If your pricing logic, security documentation and proof of results sit behind a form, the assistant building the shortlist simply can't see them. I made a related case about making your site usable by agents in WebMCP is coming for the last mile, and about why AI visibility is a leadership problem. The brief is the next step in the same argument: being visible is not enough if you're not the name in the sentence or the answer to the spec.
Consumer marketing is simply getting there first. That's exactly why I keep watching all of it.
Retail media isn't dying. The shelf it was built on is moving, out of the store and into the sentence a customer gives their agent. The platforms already know this, which is why they are fighting over the door and rebuilding the shelf inside the chat.
Marketers should know it too, whatever side of the B2B and B2C line they sit on. Fund the brand that gets you named. Fix the product truth that gets you chosen. Earn the reorder that makes you the default. And think twice before paying years in advance for a shelf your customers may stop visiting.
The click was never the point. The choice was. The choice is moving earlier. Make sure your name is in it.
Mike Bloomstine helps leaders navigate the shift from activity-led marketing to outcome-based systems. To talk through what this looks like for your business, visit mikebloomstine.com.
I help B2B and B2C teams build marketing engines that compound — strategy first, systems underneath.
