Amazon Becomes Seller, Planner and Grader

Three Inputs

Amazon's new pitch to advertisers fits on an index card. Bring a budget, pick your products, approve the creative. Amazon's AI handles the rest, from the plan to the buy to the constant reshuffling of money between sponsored ads, display, video and streaming TV.

That product is Full-Funnel Campaigns, the headline act of Amazon Ads Agent, which is the new name Amazon gave its unified ad platform at unBoxed on September 29. Signals are shared across formats, so spending early in the customer journey is supposed to make the ads closer to a purchase perform better. Results arrive in one report, with impressions, clicks and sales next to new-to-brand customers and Amazon's Long-Term Sales metric. For now, it is open only to advertisers in the United States.

Anyone who has stitched a campaign together across separate buying tools, separate teams and endless budget calls will see the appeal immediately.

A month before unBoxed, the Federal Trade Commission and 22 state attorneys general sued Amazon over how it prices the ads it already sells. According to the complaint, Amazon told advertisers for years that it ran second-price auctions, then quietly added a hidden surcharge in 2019. By 2024, the complaint says, Sponsored Products advertisers were paying their own full bid roughly 80% of the time. Amazon disputes the allegations. When Digiday asked whether the new tools would show buyers how a winning price was set, Amazon didn't commit.

So here is the arrangement on offer. The company accused of hiding how it sets prices would now decide where your money goes, execute that decision and tell you how well it worked. I find that hard to wave through as a convenience feature.

Approved at Setup, Decided at Runtime

Amazon is careful to say advertisers stay in charge. They set the goal, the campaign settings and the budget, and they approve every creative before launch. Across Amazon Ads Agent, they can also choose manual tools, let Amazon's AI optimize or mix the two.

All of that happens before the campaign runs.

Everything after belongs to the system: which format gets the next dollar, which audiences see the ad, how much goes to Amazon's own properties and how much to third-party inventory on the open web, and when the mix shifts and why. In Full-Funnel Campaigns, advertisers can't select individual channels or specific audiences at all. Amazon reports results at the format level and says it is still figuring out how to give buyers more transparency on audiences.

Experienced buyers keep a manual lane elsewhere. DVA+, which folds Sponsored Display, Sponsored TV and programmatic campaigns into one buying experience and starts rolling out in late October, has advanced settings for targeting, deals, supply sources, frequency caps and bidding. But the direction of the platform is clear. You state the objective, and Amazon works out how to reach it.

The assistant is heading the same way. Amazon says that over the coming months it will size audiences, forecast reach, recommend how budgets should be split across channels, propose targeting and diagnose performance, all through conversation. For Sponsored Products, it will suggest changes to targeting, bids and budgets that can be applied in one click. That is convenient. It also means a casual request typed into a chat window can become a spending decision, which puts access rights and approval rules squarely in governance territory.

This is the gap I keep coming back to. Organizations govern the setup, sign the approvals and file the policy. Then an agent makes thousands of decisions at runtime that nobody approved individually and few people can reconstruct afterward. Amazon has productized that gap and put a budget field on top of it.

One Company, Three Roles

Every ad platform grades its own homework to some degree. Amazon's version is unusually complete. It sells the inventory and runs the auction. In Full-Funnel Campaigns it also decides the allocation, including how much of the budget stays on its own properties. And it defines what success looks like. New-to-brand customers and Long-Term Sales may well give advertisers a broader view than last-click conversions, but they are Amazon's metrics, calculated inside Amazon's environment from shopping signals no outside party can reproduce.

Amazon's own launch announcement shows the loop in miniature. Advertisers who applied its AI targeting recommendations through the chat interface, it says, reached more than 25% more unique customers while cutting cost per impression by more than 10% on average. Amazon's recommendation, measured by Amazon, reported by Amazon. The number may well be accurate. Nobody outside Amazon can check it.

Someone will point to Amazon Marketing Cloud here. It is a capable clean room, and advertisers can bring their own data into it. It still runs on Amazon's signals, inside Amazon's walls.

Procurement teams should look closely at what they are actually buying. Amazon Ads Agent bundles inventory, data, optimization and reporting into one environment. Evaluated as a single convenience, those pieces lose their individual price tags, and alternatives become much harder to compare.

None of this would matter much if the pricing record were clean. The FTC and 22 states argue it isn't. If they are right that advertisers couldn't see how a single auction price was set, it is hard to imagine them reconstructing why an agent moved a third of their budget into streaming TV in week three.

What Agencies Keep

Full-Funnel Campaigns puts direct pressure on agency work built around campaign setup, routine optimization and manual reporting. It is hard to charge for the difficulty of moving budget between formats when the platform does it continuously and without being asked.

The specialist structure takes the hit first. Retail media, programmatic and sponsored search have often been run by separate teams with their own tools and reporting conventions. One campaign that crosses all three boundaries weakens the case for keeping them apart. Holding companies spent years building those practices. Amazon just collapsed them into a single setup screen.

Amazon's launch release includes an endorsement from dentsu's head of commerce and retail media, who says the chat interface lets his strategists spend less time on execution and more on insight and creative thinking. That is an accurate description of what is happening. It is also a strange thing for an agency to applaud inside a platform's press release.

What remains is the work above the platform. Someone has to decide whether the client's business objective has been translated into the right campaign goal, whether Amazon deserves the next dollar more than another retailer, a search engine, a social platform or the open web, and whether Amazon's reported results survive an experiment Amazon didn't design. Creative judgment belongs in the same category. Amazon can generate video and display ads from existing product pages, but producing an ad faster doesn't tell a brand which message it should carry into the market, or whether a hard-selling treatment is quietly eroding its position.

There is an uncomfortable wrinkle in that argument. The obvious candidate for the advertiser's independent check on Amazon is the agency. But many of the same holding companies run principal media books, buying inventory and reselling it to clients at a margin the client never sees.

An independent decision layer can't sit with anyone who profits from where the money lands. That applies to Amazon. It applies to the agency, too.

Who Checks the Math

Amazon provides format-level reporting for Full-Funnel Campaigns and says it is building explanations into Amazon Ads Agent that tell buyers why it recommends particular targeting or budget moves. Both are welcome. Neither is an audit.

An advertiser should be able to see how much money moved between formats, when it moved and what evidence prompted the move. It should be able to see which supply sources were used, what fees were charged and how each automated decision affected reach, frequency and price.

Without that record, continuous optimization can't really be audited. An explanation saying the system increased streaming TV spend because it expected incremental sales tells you what happened. Whether the forecast deserved belief is a separate question, and answering it takes evidence buyers can hold up against observed results and outside measurement.

Amazon is right that all of this opens sophisticated buying to advertisers who never had a programmatic team. That is also what worries me most.

The advertisers most likely to hand Amazon a goal and a budget and walk away are the ones least equipped to run a holdout test, question a Long-Term Sales figure or notice the mix drifting toward inventory that served Amazon better than it served them. The FTC complaint counts more than 500,000 small and mid-sized businesses among the advertisers it says were hit by the surcharge.

Large advertisers will benefit too, through lower coordination costs and specialists freed from repetitive campaign work. But once a brand hands Amazon a goal, a budget and approved creative, Amazon turns that brief into a media plan and keeps rewriting it as results come in. The agency or in-house team stays accountable for an outcome it no longer controls.

That only works if someone outside the platform keeps the decision layer. Someone has to judge whether Amazon's objective matches the company's, whether the platform deserves the budget it recommends and whether reported success holds up under outside scrutiny. Those judgments can't be handed to the system that sells the inventory.

As platforms take over more of the execution, the client relationship will stop belonging to whoever can operate the buying interface. It will belong to whoever can frame the decision, test the evidence and tell the platform no.