Amazon Just Handed Your Media Plan to an Agent. What It Will Optimise Is Whatever Your Listings Already Say

Amazon Just Handed Your Media Plan to an Agent. What It Will Optimise Is Whatever Your Listings Already Say

By Stephen Honight, Founder of Lmo7

On 29 September, at its unBoxed conference in San Francisco, Amazon renamed its unified ad platform Amazon Ads Agent and launched Full-Funnel Campaigns and DVA+. In Full-Funnel Campaigns you supply the products you want to sell, the creative and a budget. Amazon’s AI decides how to execute and keeps optimising toward outcomes that include new-to-brand customers and Amazon’s long-term sales metric. That is Ronan Shields’ account in Digiday, published on 29 September 2026.

The machine now runs more of the buying. My sense is that the useful question is what it optimises against, because that is where your own work still counts. If you want the practical part first, the split of what to hand the agent and what to keep human is near the end, along with the one number I would agree before launch.

A note on sources. Amazon’s own announcement page for these launches could not be read at the time of drafting, so this piece relies on Digiday and on Laurie Sullivan’s report in MediaPost, also dated 29 September 2026. I read both in full and name which one each fact comes from.

This is the third agent in a set. The shopper side is covered in Your Amazon Ads Are Already in the Conversation and the seller side in Amazon Just Put Seller Central Inside Claude. This one is the buyer-side agent that runs the media buying.

What actually launched

MediaPost reports that Amazon Ads now offers three streamlined campaign types under the Amazon Ads Agent name: Sponsored Ads, DVA+ and Full-Funnel Campaigns. The platform replaces the former Ads Console and Amazon DSP experience and includes a chat experience in natural language. MediaPost adds that the announcement follows a year-long initiative to merge siloed ad tools into one AI-powered ecosystem.

DVA+ is the programmatic side, meaning the automated buying of display, video and audio placements. Digiday says it consolidates Sponsored Display, Sponsored TV and programmatic display, video and audio buying. In simplified mode the advertiser selects an objective, budget and creative, and Amazon handles targeting and optimisation. Advanced settings keep control over targeting, deals, supply sources, frequency and bidding. Kelly MacLean, vp at Amazon Ads, called it “human-led AI supported” in Digiday, and told MediaPost “With DVA+, we have brought the power of programmatic to advertisers of all sizes”. MediaPost adds that DVA+ is powered by Brand+ and Performance+.

Full-Funnel Campaigns is the bigger change to how a plan gets built. Per Digiday it combines sponsored ads, display, video and streaming TV within a single campaign, and can include third-party open-internet inventory. MediaPost says it runs on Amazon.com and Amazon mobile (home page, search, display) as well as third-party properties, and shares signals in real time across formats so upper-funnel budgets support lower-funnel performance. Results are measured with standard metrics plus new-to-brand and long-term sales.

Two details from MediaPost matter for planning. Full-Funnel Campaigns requires a minimum daily budget of $100 and is currently available to all advertisers in the United States. Nothing I read covers timing elsewhere, so I would treat UK availability as unknown. For a challenger brand, that $100 floor decides whether this is a test you can run properly.

The platform also includes a conversational layer. MediaPost says it is backed by what Amazon describes as trillions of shopping, browsing and streaming signals. New capabilities include media planning with data-backed budget recommendations, comparing incremental reach across channels, describing an audience in plain language to get targeting recommendations, and asking performance questions to diagnose trends and benchmark against peers. More are promised over the next year. Digiday describes Amazon expanding what began as a conversational interface into an AI layer that spans planning, campaign creation, optimisation, measurement and support.

What an AI buying layer optimises against

Amazon has not set out the model’s inputs in anything I read, so what follows is my reading and not something Amazon has said.

The stated outcomes are new-to-brand customers and long-term sales (Digiday), and MediaPost says signals are shared in real time across formats. An agent working toward those outcomes can only learn from what shoppers do. Do they click through, and do they buy once they land. My view is that retail signals like these are what it steers by.

Those signals come from the listing. The title influences whether the click happens. Bullets, A+ content and images decide whether the shopper understands the product. Reviews, price and availability weigh on whether they buy. Over time the ASIN builds a behavioural history from all of it. An agent can find more people to send to a page. It has no say in what happens when they arrive.

MediaPost’s detail on creative makes the dependency plain. Advertisers can use Amazon’s built-in creative tools to generate images and TV-ready video from existing product pages. The product page is now the source material for the ad as well as the destination. A thin page produces thin creative.

So automation scales what is already there. A listing with a weak main image, generic bullets and a handful of reviews is now easier to push into display, video and streaming TV. The reach is easier to buy and the page at the end is unchanged. Shoppers who land and do not buy send the system a signal, and I would expect the agent to learn from that, but only after the budget has paid for the lesson.

Content and eligibility therefore set the ceiling on ad efficiency as well as on organic recall. By eligibility I mean the plain conditions: in stock, priced where the market will bear it, claims correct and compliant. No agent fixes an out-of-stock ASIN.

One reading of this news is that the agent runs the buying, so a brand needs less help. I have a different view, which is that the agent runs the buying and the advantage moves to the inputs. Amazon says Full-Funnel Campaigns is open to all advertisers in the United States (MediaPost), so the agent itself will not separate one brand from another. Content, data and tracking will, and that is the work we do at Lmo7.

What you give up, and what stays with you

Digiday is clear that buyers give something up. In Full-Funnel Campaigns advertisers approve creative but cannot select individual channels or specific audiences. Amazon provides visibility at format level. Amit Bhattacharyya, vp of full-funnel products, AI & modeling at Amazon Ads, gave the rationale as “What if we made it really simple, and yet very powerful for advertising?” On audiences he told Digiday “We are figuring out how to give them more transparency on the audience side.”

I understand the appeal. A small team with one person running ads has good reasons to want simple. What stays with you is creative approval and the budget, and in DVA+ the advanced settings over targeting, deals, supply sources, frequency and bidding.

Digiday names three areas for marketers to question:

• Whether you can audit what the agent does with budgets and creative.
• How well it works with third-party data, identity and measurement providers.
• Whether fees, supply sources and auction mechanics stay clear.

One piece of context, kept short. Digiday notes that in August 2026 the US Federal Trade Commission and 22 state attorneys general sued Amazon over alleged surcharges in its ad auctions, which Amazon disputes, and I am drawing no conclusion on the case. The detail that matters for an operator is that when Digiday asked whether the new tooling would show buyers how individual winning prices were determined, Amazon did not commit to that level of auction transparency.

My recommendation would be not to sit this out. The mechanism above says the inputs matter most, and the inputs are yours. Keep an audit line instead: your own record of what was spent, on which formats, behind which products, and what came back.

Efficiency targets

The target you hand an agent becomes its constraint. The reporting I read does not say whether Full-Funnel Campaigns accepts a ROAS target (return on ad spend, the revenue each pound of ad budget brings back), and DVA+ advanced settings cover bidding, so this applies wherever you can set one. Set it too tight and the system buys only the impressions it expects to clear the bar. It buys less volume while the reported efficiency looks excellent. My recommendation would be to judge the agent on incremental volume at an acceptable efficiency, and to agree what acceptable means before launch, with your margin in front of you.

What to hand over and what to keep

Here is how I would split the work.

What I would hand to the agent:

• The objective. Digiday and MediaPost name new-to-brand customers and long-term sales as outcomes, alongside standard metrics. Decide which you want it to lead with.
• The guardrail. The efficiency target above, set with your margin in view.
• The ASINs that deserve budget, which is a shorter list than the catalogue.

What I would keep human:

• Which products are ready for traffic. Page quality, stock, price and reviews are your call. You know about the stock delay on the hero product. The agent may not.
• The claims on the page. Product pages feed the generated creative (MediaPost), so an inaccurate claim can be copied into an image or a TV spot. Someone owns accuracy and compliance.
• The approval line. Digiday says advertisers approve creative in Full-Funnel Campaigns. Decide who signs off and against what.
• The audit. Amazon provides format-level visibility (Digiday). Keep a weekly record of what the agent did and what it spent.

A proof point, with a caveat

This next result is organic and it is not an ads result. In a 60-day AI search engagement for Trip Drinks, a current Lmo7 client, average position across major LLMs moved from 7th to 3rd and AI referral traffic rose 33%. Both figures come from Lmo7’s own LLM visibility tracking across that 60-day window, as published in our Trip Drinks case study. I am not claiming it predicts how an ads agent will perform.

I include it for the input. The content and product information that help an AI system recommend a product are the same material the ads agent now reads and builds creative from. Improving that layer moved an outcome I can measure. My expectation, and it is only an expectation, is that a stronger source page gives the ads agent better material to work with. I have no ad-side result to show for that yet.

What I would do next

Here is the order I would take it in.

1. Settle which ASINs are ready for traffic, and fix the rest. Owner: your Amazon or ecommerce lead, with Lmo7 on the content. Commercial reason: the agent scales what is on the page, so the page is where efficiency is won first. For a challenger brand I would start on the Lmo7 Challenger tier by default. Amazon Content Optimisation is £750 a month, so the agent optimises a listing worth scaling. Amazon Ads is £1,000 a month for managed ads with weekly optimisation. The Full Amazon Stack, which combines the ops, content and ads modules, is £2,500 a month.

2. Build the audit line before budget goes live. Owner: your analytics or finance lead, with Lmo7. Commercial reason: Amazon provides format-level visibility in Full-Funnel Campaigns and did not commit to auction transparency (Digiday). Lmo7 DaaS is £250 + VAT per month per marketplace and gives direct API access, so your own team can audit what the agent did.

3. Agree one governance model if you run several brands. Owner: your group ecommerce or commercial director. Commercial reason: when each brand team hands its own budget to an agent, objectives, guardrails and approval lines can drift apart. Lmo7 Enterprise workshops start from £5,000 and cover multi-brand governance of agent-run budgets.

Stephen Honight is the founder of Lmo7, the AI-native agency helping consumer brands win in AI-powered discovery and agentic commerce across Amazon and D2C. Lmo7 works with brands including Trip Drinks, Veloforte, Brown-Forman, Haleon, Pelotan and Symprove.