What changed: from Ads Console to Ads Agent

For years, advertising on Amazon meant opening the Ads Console and adjusting campaigns, bids, keywords and budgets by hand. At unBoxed 2026, Amazon changed the game: the console gave way to the Amazon Ads Agent, a conversational, agentic assistant. You tell it what you want, and it plans, buys, optimizes and reports — including more opaque inventory such as Branded Conversations.

It's convenient. The catch is that when the machine decides for you, the question stops being "how do I set up the campaign?" and becomes "how do I check what the machine bought?". Independently verifying what the agent did stops being a luxury and becomes a control function.

The problem: the agent grades its own homework

The Ads Agent is good at execution. But it optimizes for Amazon's metrics — clicks and ACoS — and it has a structural blind spot: it doesn't know your product cost, your fees or your taxes. In other words, it doesn't know whether the sale it bought was profitable.

The result is a silent risk: a campaign with a "pretty" ACoS can, after real costs, be losing money. Without an independent check, you accept what the agent bought blindly — and only find the hole at the end of the month.

What the Ads Agent Audit is

The Ads Agent Audit is your second set of eyes on Amazon's agent. It doesn't compete on execution (against a free, native agent, that wouldn't make sense). It covers the layer Amazon doesn't provide: the truth.

In one sentence: it compares the goal you set for the account with where the money actually went, and closes the books on real net profit (COGS + taxes + fees). All of it read-only: the Audit reviews and reports — it never touches your account or your campaigns.

Step 1 — set the account's goal

Without a goal, there's no divergence to measure. So the first step lives in Account Settings (System → Settings). In the "Goal for this account" card, you choose Grow sales, Lower customer acquisition cost (CAC) or Defend the brand. That's the yardstick every dollar of spend is measured against.

Step 2 — read the Divergence Report

With the goal set, the Ads Agent Audit screen (Intelligence menu) builds the Divergence Report. From top to bottom, it answers:

  • The verdict, tied to the goal: how many critical points exist where the agent's spend didn't serve what you asked for.
  • Key numbers in cards: Ad spend, ACoS (against your target), Real net profit, Waste (spend with no sales) and Cost per order. Note: net profit is the number Amazon doesn't show you.
  • "Where spend diverged from the goal": each divergence ranked by severity (critical, warning, note), with the number behind each one and what to do. In the example: $901.89 in spend with no sales (critical), 10 campaigns above the ACoS target and overall ACoS above target.
  • "What can't be audited": the honest limits — Branded Conversations and part of the auction are opaque on Amazon; without branded search term classification, that slice is declared as a limit, never made up.

Step 3 — ask Vision

You don't even need to open the screen: Vision, EcomVision's AI, answers the Audit in plain English. Ask "did the Ads Agent spend on what I asked for?" or "are my ads actually profitable?" and it brings back the verdict, the divergences (critical ones first) and the real net profit — and, if the account goal is missing, it walks you through setting it.

For sellers

If you run your own account, the Audit answers the questions Amazon's dashboard leaves hanging:

  • "Is the agent working for me?" — see whether spend followed your goal, or went wherever was easiest (defending the brand when you wanted to acquire new customers, for example).
  • "Is this good ACoS actually profit?" — the math closes on net profit, after cost, fees and taxes. No end-of-month surprises.
  • "Where am I burning budget?" — waste (spend with no sales) and off-target campaigns show up in dollars, ready to take back to the agent or fix by hand.

For agencies

If you manage multiple accounts, the Audit is ammunition for client relationships and sales — the Software with a Service (SwaS) angle:

  • Proof of value with a profit number: instead of an ACoS screenshot, you bring the client a Divergence Report showing the budget went to the agreed goal and that real profit was left over.
  • Scalable governance: read-only and standardized, the Audit becomes an audit routine that runs the same way across every account — no manual spreadsheets.
  • Hard conversations, backed by data: when Amazon's agent spent poorly, you have the verdict and the numbers to justify the adjustment, not just a hunch.

Honesty as a differentiator

Part of Amazon's inventory is opaque by design — Branded Conversations and parts of the auction can't be audited by anyone outside Amazon. The Audit flags this explicitly in the limits section instead of pretending to measure what it can't. The same goes for the branded-term slice: until the data exists, it's declared as a limit, never estimated in the dark. That standard — audit, never invent — is what makes the number reliable enough to take to a decision (or to a client).

Update note: the Amazon Ads Agent is still rolling out (in beta in the US since January 2026), and names and breakdowns may change in Amazon's interface. The Audit evolves with it: audience and branded-term breakdowns will be added as the new Ads API exposes that data.

Sources