The first agents advertisers are being handed to spend their budgets live inside the platforms that sell the inventory. That is a sensible place to start and a poor place to stop. Buyers need an agent of their own, one that no seller owns.
This is an opinion, and the argument is mine. The facts below come from two documents I read this week.
The promise is judgment at scale
Matthew Jacobs, Head of Agentic Technologies at IAB Tech Lab, wrote this week that “Agentic advertising allows buyers high-judgment decision making at scale for the first time.” His case is persuasive. A buyer today logs into a dozen interfaces, falls back on the two or three they know by heart, and never considers the options that would have fit the brief better, because reaching them costs too much administration. An agent that can weigh thousands of options against a brief removes that cost.
Jacobs is also clear about the condition. In his post he says agents “need a shared way to understand what a publisher is offering, communicate a buyer’s requirements, and establish what they are authorized to do.” Without common standards, he writes, the industry risks recreating the fragmented workflows it has now. That is the gap we described in Ad Tech Can’t Agree on One Language for AI Buyers.
What is shipping instead
What is shipping this month is agents attached to individual platforms. On September 29, Amazon Ads renamed its combined DSP and ads console Amazon Ads Agent, with Full-Funnel Campaigns and DVA+ inside it. Amazon says its AI uses signals from shopping, browsing and streaming to place ads across Amazon properties and the open internet, and that advertisers can choose manual tools, AI optimization or a mix. Our own coverage in Agentic AI Enters Ad Buying, One Workflow at a Time tracked the same pattern at smaller scale.
I do not read this as a problem with any one company. Every large platform will build the same thing, and each should. A platform agent knows its own signals and its own inventory better than anyone. The structural point is narrower. An agent that works for you but is built, hosted and updated by a seller can optimize well inside that seller’s world. It cannot be the neutral party that compares that seller against the others.
The counter-argument
The strongest objection is that platform agents already hand buyers the controls. Amazon’s announcement says advanced settings let experienced programmatic buyers set targeting, deals, supply sources, frequency caps and bidding strategies. Control over levers is real, and it matters.
But control inside one platform answers a different need from comparison across platforms. A buyer with full control of every setting in one console still cannot see, from inside that console, whether a dollar would have done more somewhere else. Jacobs’s own premise is that buyers fall back on familiar interfaces because comparing is costly. A platform agent that makes the familiar interface better makes the familiar choice easier still.
What a buyer-side agent needs
The pieces for an independent layer are arriving. Amazon announced an Amazon Ads MCP Server (Lite) that, per its unBoxed 2026 recap, “simplifies how agents interact with Amazon Ads”. That matters because it means an agent the buyer controls can call a seller’s tools instead of living inside them. Standards work at IAB Tech Lab is aimed at the shared vocabulary Jacobs describes.
I would ask any advertiser building or buying an agent to settle four things before the first budget moves.
- Who owns the agent’s instructions and its log of decisions, and can you export both.
- Which sellers the agent can call, and whether it can call one you add next quarter.
- Whether its reporting uses your definitions or the platform’s.
- Which spend it is allowed to move without a person approving it.
Those answers decide whether the agent works as your buyer or as a very helpful sales channel. Platform agents are good tools and I expect to use them. What I would not do is let them become the only agents in the room.
What marketing leaders should do
Run platform agents where they have the best data, and keep a record of every recommendation they apply. Fund a small buyer-side pilot that compares at least two sellers against one brief, using your own definitions of reach and outcome. Join the standards conversation, because the vocabulary being written now will decide how easily you can leave a platform later.
Agentic buying will reward the advertisers who keep the comparison in their own hands.
Source: IAB Tech Lab
