Amazon Ads Agent is three control modes, not one automation setting
A comparison card separates Amazon Ads Agent’s manual, assisted, and delegated modes from their availability states.
Amazon Ads Agent now puts several advertising paths under one product name. Amazon says advertisers may build campaigns manually, delegate optimization to Amazon AI, or combine the two approaches across sponsored ads, Full-Funnel Campaigns, and Display, Video, and Audio (Amazon Ads).
That shared surface does not make those paths operationally equivalent. For buyers, the useful unit of comparison is the control mode attached to a campaign type and release state.
Read the product as a set of control modes
The source material supports three broad modes:
- Manual control: the advertiser configures and operates the campaign.
- AI-assisted control: the advertiser uses AI within a workflow while retaining some configuration choices.
- Delegated execution: the advertiser assigns optimization work to Amazon AI.
These labels are our interpretation of the choices Amazon documents, not Amazon’s formal taxonomy. Amazon explicitly describes manual building, delegated optimization, and mixed operation (Amazon Ads). AdExchanger separately reports that buyers can retain an advanced, traditional DSP-style workflow or use DVA+ (AdExchanger).
The distinction matters because “available in Amazon Ads Agent” does not, by itself, identify who configures the campaign, what the system may change, or where a person approves an action. Those are diligence questions derived from the documented product choices; the supplied sources do not answer each one.
The SGL comparison card
Use one row per campaign type and operating mode rather than one row for the entire platform.
| Field | What to record |
|---|---|
| Campaign type | The specific campaign or inventory path being evaluated |
| Human control surface | The settings, constraints, or workflow the buyer directly controls |
| Delegated action | The action assigned to the platform, stated as narrowly as the evidence permits |
| Approval point | The moment a person reviews or authorizes the action, or “not documented” |
| Availability state | Available, rolling out, in development, described as future, or unknown |
A source-bounded first pass for Amazon would look like this:
| Campaign or path | Control mode | Delegated action | Approval point | Availability state |
|---|---|---|---|---|
| Sponsored ads and other manually built campaigns | Manual | None established by the excerpt | Not applicable from the supplied evidence | Included in the unified surface |
| Campaigns using Amazon AI optimization | Delegated or mixed | Optimization | Not documented in the supplied excerpts | Included as a documented operating choice |
| Full-Funnel Campaigns | Mode not fully specified in the excerpt | Not documented precisely enough to record | Not documented | Available in the US |
| DVA+ | AI-assisted or delegated mode suggested, with the exact boundary unresolved | Not documented precisely enough to record | Not documented | Begins rolling out in late October 2026 |
| Planning and forecasting | Unresolved | Future capability remains unspecified | Not documented | In development according to AdExchanger |
Amazon documents Full-Funnel Campaigns as available in the US, says DVA+ begins rolling out in late October, and places additional conversational capabilities over the coming months (Amazon Ads). AdExchanger reports that planning and forecasting remain in development and describes fuller automation as a future possibility (AdExchanger).
The table deliberately leaves gaps. Neither excerpt specifies every approval point or the exact action boundary for each campaign path. Filling those cells from the “agent” label would convert product positioning into an unsupported control claim.
Compare authority before feature names
A competing platform may use different language for similar operating choices, or similar language for materially different authority. The card keeps the comparison anchored to five questions:
- Which campaign type is in scope?
- Which controls remain directly available to the buyer?
- Which actions can the system take?
- Where, if anywhere, does human approval occur?
- What is the capability’s present release state?
This is a proposed procurement frame, not a claim about observed Amazon Ads Agent behavior. A buyer could populate it from product documentation, contract terms, demonstrations, and controlled evaluation results, while labeling each kind of evidence separately.
The release-state field is especially important here. A capability available in the US, one beginning a rollout later in the month, and one still described as in development do not offer the same procurement certainty. The supplied sources establish those different descriptions, but they do not establish account-level access, operating performance, or rollout completion.
One name, several evaluation objects
Amazon’s unified branding makes navigation simpler to describe, but the evidence still presents multiple control arrangements and timelines. The procurement object is therefore better treated as a tuple:
campaign type + control mode + delegated action + approval point + availability state
That tuple is an SGL-derived evaluation device. It prevents a platform-level label from standing in for details the sources leave open. Teams applying the same method to their own agent systems can carry these fields into a concrete build review, where each delegated action can be tied to an explicit control surface and release state.
For Amazon Ads Agent, the immediate conclusion is bounded: the announced surface spans manual, mixed, and delegated operation, while individual capabilities occupy different release states. Any stronger claim about approval mechanics, account availability, or end-to-end automation awaits more specific evidence.