Amazon Ads Rebrands DSP as Agentic AI Platform at unBoxed

Full analysis

Amazon killed the DSP brand. At its unBoxed event, the demand-side platform advertisers used to buy programmatic inventory gets replaced by DVA (Display, Video and Audio), merged with the Amazon Ads console under one chatbot called The Amazon Ads Agent. The default setup is now two AI campaign types, Performance Plus for conversions and Brand Plus for branding. The old hands-on DSP is a toggle away in "advanced settings." Amazon is betting that the way you buy ads is a conversation, not a console.

What's actually being decided here: nothing, for Amazon. This shipped. The decision belongs to everyone who buys Amazon inventory, and to Amazon's rivals who now have to answer the same "just tell the agent what you want" pitch. Easy to undo on your side: you can ignore the agent and live in advanced settings. Hard to undo on Amazon's side: they've now told the market the console era is over. What sets the clock is the next few earnings cycles, when we find out if this pulls new budget or just reshuffles the spend Amazon already had.

The Skeptic. Amazon has "simplified" this platform before. The Sizmek ad-server buy in 2019 was supposed to unify things and instead bought years of fragmentation that this rebrand is now cleaning up. Calling a chatbot layer "agentic" is a marketing choice. If the bidder underneath is the same bidder, and the data pipes are the same pipes, your return on ad spend does not move because the login screen got friendlier. The small-advertiser Full-Funnel pitch is the oldest story in the platform playbook: push SMBs into streaming TV with auto-made video. Everyone has tried it. Brand metrics always disappoint. For this to matter, the agent has to pull budget out of Google and Meta, not just consolidate spend Amazon already owned.

The Enterprise Buyer. No agency trading desk is going to hand budget authority to a chatbot. The people who spend real money on Amazon DSP are holding companies and in-house teams with reporting obligations, log-level data needs, and finance departments that want to know why every dollar moved. They will live in "advanced settings" forever. That is the migration trap: Amazon now maintains two full interfaces and two sets of API contracts, one for the demo and one for the customers who actually spend. The agent is aimed squarely at the long tail, the advertiser too small to staff a trader. That's a real market. It is not the market that pays Amazon's ad-revenue line today.

The Researcher. The genuinely interesting claim is that the agent "retains context across campaigns." Ask whether that context does causal work or is decoration over the same bidder. Amazon's signal is unmatched, purchase intent plus streaming behavior plus Alexa. But Performance Plus and Brand Plus are multi-objective problems with a murky reward: what is the agent optimizing when it balances a conversion goal against a branding goal inside one campaign? Nobody has shown that carrying session memory improves outcomes versus a stateless setup wizard. Kelly MacLean, Amazon's VP of science, engineering and product, said planning and forecasting are still in development. That is the admission that matters. The autonomous loop she's describing is not closed. The pieces that would make it a real optimizer are the pieces not yet shipped.

The Compute Pragmatist. Every agent interaction that touches creative generation, media planning, and bid optimization is a chain of model calls, not one API hit. Retaining cross-campaign memory across millions of live campaigns is persistent session-state infrastructure that costs real money to run. A traditional DSP API call is cheap. An agentic campaign setup is an order of magnitude more expensive to serve. Amazon runs this on its own Trainium and Inferentia silicon, so it eats that cost better than a rival renting NVIDIA capacity. But there's a reason the default steers you toward the automated tiers: automation that Amazon controls is automation Amazon can price. The console was a cost center. The agent is a margin lever.

The Safety Lens. An agent that writes video and executes a media plan with minimal human checkpoints can spend an advertiser's money on streaming inventory before anyone reviews it. When the agent places a brand-unsafe auto-generated video at scale, who is accountable? The creative generator, the media planner, or the bidder? The "advanced settings" toggle being the only off-ramp for human control is thin governance. Auto-made video next to brand-sensitive inventory needs moderation tooling that Amazon has not described publicly, and regulators are already circling autonomous systems that move money without a human in the loop.

Where the council splits. Two real disagreements. First, the Researcher and the Skeptic both suspect this is a skin over the same bidder, while the Compute Pragmatist thinks the machinery underneath genuinely changed and got much more expensive to run. Those can both be true: new plumbing, same output quality. Second, the Enterprise Buyer says the paying customers never leave advanced settings, while Amazon's whole pitch assumes the automated tiers become the default that grows the pie. That tension is the decision. If the agent only converts existing Amazon spenders into a cheaper-to-serve automated flow, this is a margin story. If it drags budget off Google Performance Max and Meta Advantage+, it's a growth story. The announcement can't tell you which, and Amazon is betting you'll assume the second.

What it hinges on. One belief: does an auto-generated, auto-optimized campaign beat a human-set-up campaign on the metrics buyers actually pay for? Google Performance Max, the closest comparison because it's the same "let the machine allocate" pitch across a walled garden's inventory, took two years and still leaves sophisticated buyers demanding channel-level controls back. Amazon is walking the same road. Before you move budget, run the automated tier against a human-built campaign on the same objective, same flight, and split test. Look at incremental conversions. Ignore the dashboard's self-graded number.

Prediction: By Amazon's Q4 2026 earnings call in late January or early February 2027, Amazon will report advertising revenue growth that is roughly in line with or below its trailing four-quarter average, showing the agentic rebrand did not produce a step-change in ad spend within its first full quarter.

Confidence: Medium. UX rebrands consolidate existing spend faster than they win new budget.

Why: The rebrand launched at unBoxed and defaults advertisers into automated tiers, but Amazon's own VP Kelly MacLean says the planning and forecasting tools that would close the autonomous loop are still in development, so the product buyers would need to trust with net-new budget isn't finished yet. The paying core, agency trading desks and large in-house teams, will stay in "advanced settings," meaning the near-term effect is cheaper-to-serve consolidation of spend Amazon already had, not budget pulled from Google and Meta. Google's Performance Max took years to earn sophisticated buyers' trust and still faces demands for manual controls; a one-quarter-old chatbot layer over the same bidder won't move a $50B+ ad business off its existing trajectory that fast. The opposite outcome, a visible acceleration in ad growth attributable to the agent, would require SMB streaming adoption to convert immediately, which every platform's history says it doesn't.

Revisit by 2027-02-15: We're right if Amazon's Q4 2026 ad revenue growth rate is at or below its trailing four-quarter average. We're wrong if it comes in above that average.