Every year, Amazon uses its unBoxed keynote to signal where its ads business is heading, and this year’s signal was hard to miss. It didn’t come out of nowhere, though. Last year’s unBoxed unified Sponsored Ads and Amazon DSP under one Campaign Manager. Ads Agent already existed tucked inside DSP and AMC. And in February, Amazon’s Ads MCP Server opened for beta. This year’s thirteen announcements are what happens when those three threads get pulled together in one story: shoppers discover products through conversation instead of search, advertisers buy through one agent powered platform instead of several consoles, and insights turn into action without anyone writing SQL.
Amazon organized all of it around three themes: Conversations and Commerce, Increased Efficiency, and Insights to Impact. Here’s what’s in each one and what it actually means if you weren’t in the room.
One naming note is worth clearing up first, since it touches almost everything below. Amazon merged Sponsored Ads and Amazon DSP into one platform called Amazon Ads Agent, with three campaign types underneath it: Sponsored Ads, Full Funnel Campaigns, and a new Display, Video & Audio campaign type, DVA, which comes in two flavors. DVA+ hands optimization to Amazon’s AI. Advanced Settings keeps full hands-on control, with an AI co-pilot layered on top.
The Conversation Doesn’t End at the Search Bar
Amazon’s theme: Conversations and Commerce
Branded Conversations is the clearest signal of where Amazon thinks shopping is headed. Launching in general availability in January 2027 in the US, it starts from a Sponsored Brands Collections ad and turns a single question-and-answer exchange into a multi-turn dialogue inside Alexa for Shopping. A shopper comparing electronics or building out a skincare routine can keep asking follow-up questions, and the AI keeps answering from the brand’s own catalog, product guides, compatibility information, imagery, and video, not a generic script. An optional “conversation brief” lets advertisers steer what the AI emphasizes. It’s priced to be low-risk to test: it turns on automatically for Sponsored Brands Collections campaigns, and advertisers pay only for the first prompt click. Amazon says shoppers who use Sponsored Prompts today convert 48% more often, and Branded Conversations is built on that same foundation.
For brands, that shifts extra weight onto content operations. The quality of the conversation is only as good as the quality of the content feeding it, so keeping product guides, compatibility details, and imagery current stops being a back-office task and starts being the material an AI draws from directly.
Conversational prompts in Responsive eCommerce Creative, available now across 18 locales, bring that same conversational format off Amazon entirely, into ads running across the open internet.
Review Requests, in open beta through the end of October in the US for ASINs under 1,000 reviews, is a new campaign type that asks verified buyers to leave a review on Amazon’s highest-traffic pages. It only charges when a shopper clicks to submit one, and in closed beta it tripled weekly reviews, a strong lever for newer or lower-review products trying to build trust fast.
Sponsored Services rounds out this theme, matching shoppers in the US with local service providers based on what they just bought. It’s a narrower play, built for local service businesses rather than retail brands and agencies.
Every new format here becomes another line item to plan, pace, and report on. Pacvue integrates new Amazon ad types as their APIs become available, so brands manage them alongside everything else instead of in a separate console.
One Agent, Three Ways to Buy
Amazon’s theme: Increased Efficiency
Amazon Ads Agent is the biggest structural change of the week: Sponsored Ads and DSP now sit inside one connected buying experience, with a single AI interface for planning, building, and optimizing campaigns.
Full Funnel Campaigns, in open beta in the US now, lets an advertiser set one budget, add products, and let Amazon’s AI handle targeting, creative generation, and pacing across Sponsored Ads and DVA inventory at once. It runs alongside a brand’s existing campaigns, so teams can test it without disrupting anything already working. That’s a deliberate design choice: one algorithm managing an entire budget with no segment-level control works well for a smaller or newer advertiser who wants a fast, guided way into full-funnel media. An enterprise team running a complex, multi-market portfolio may need a different kind of visibility down to the segment level, which this setup doesn’t offer.
DVA+ and Advanced Settings give that same full-funnel inventory two different ways in, depending on how hands-on a team wants to be. DVA+ hands optimization to Amazon’s AI against an objective a team sets. Advanced Settings keeps the full programmatic toolkit and full manual control, now with an AI co-pilot alongside it. Amazon built this to fix something that’s been building for years: 77% of traders say they feel overwhelmed by campaign complexity, and teams lose more than 10 hours a week to manual campaign management.
That’s already familiar territory for Pacvue’s enterprise customers. Pairing Prime Video Ads for pre-event demand with Pacvue-managed DSP retargeting drove Danish Endurance a 740% increase in CTR and 20% more incremental reach, and Colgate, working with GroupM, saw 228% ROAS growth during Prime Day and 75% ROAS growth in the lead-in period using Pacvue’s AMC-informed targeting.
Media Planning and Natural-Language Analytics, expanding through beta over the coming months in the US, rounds out this theme with two new agent skills: one that builds a media plan in minutes, and one that turns performance data into insights through plain conversation instead of a dashboard.
Amazon’s goal with Ads Agent is to collapse fragmented buying into one AI-powered surface. Pacvue has been building toward the same thing, with unified, agent-driven buying across a brand’s full retail media footprint.
Opening Up the Machinery Behind the Agent
Amazon’s theme: Increased Efficiency, continued
For technology partners building on Amazon’s infrastructure, two more announcements stood out. The Amazon Ads MCP Server is adding a lighter consumption model: instead of an AI agent having to hold all 115-plus individual tools in memory, a new Light MCP endpoint exposes just three meta-tools that search for and invoke the right one per task, alongside the full tool set that’s already there. Amazon says that cuts the context an agent needs to carry from roughly 98,000 tokens down to under 5,000, freeing up room for partners running multi-retailer, multi-tool agents to bring their own expertise alongside Amazon’s.
Sitting on top of that same MCP server isExpert Tools: an AMC SQL Generator that turns a plain-language question into a validated, ready-to-run AMC query. It will be priced at cost when it reaches general availability in 2027. It opens AMC up to teams that don’t have a dedicated analytics resource, and Pacvue has already integrated it into Pacvue Agent for AMC query generation. Writing the SQL is only part of the work. The harder part is knowing which question to ask, and turning the answer into an audience, an activation, and a measured result without that becoming its own separate project, which is the same workflow Pacvue built this tool into rather than treating it as a standalone step. Agencies already using Pacvue Agent this way have seen it pay off: Itsumo saved more than 200 hours and grew new-to-brand customers 76% building AMC audiences through Pacvue Agent instead of writing SQL by hand.
When Insight Doesn’t Wait for a Report
Amazon’s theme: Insights to Impact
Geographic Insights and Activation, expanding globally for brands selling on Amazon, brings ready-made sales-penetration data from the Amazon Sales Index directly into DSP location targeting. A team can target low-coverage regions to find new customers, or high-coverage regions to reinforce loyalty, in a few clicks, without building a custom AMC query first.
Publisher Insights in AMC, in beta now in the US and Canada, brings show, genre, and content-category data into the AMC workflow, so a team can see which content is actually driving demand without writing SQL or exporting to another tool.
The redesigned Amazon Marketing Stream V2 is a quieter announcement with real downstream impact. Data will land in a partner’s own S3 bucket in Parquet format, covering every metric in the reporting API, roughly every 15 minutes instead of hourly, with a solid chunk of historical backfill available as soon as a partner subscribes. For anything time-sensitive, like reacting inside a live event window, that’s a materially faster feedback loop than what teams have worked with until now.
Amazon is pushing insights closer to the point of activation here, the same move Pacvue has been making. Pacvue already supports Geographic Insights in DSP line-item setup and uses Amazon Marketing Stream data to power capabilities like Live Ad Momentum and dayparting.
Where This Leaves Advertisers
Taken together, this year’s unBoxed keynote confirms a strategy Amazon has been building toward for a while: unify the buying experience, hand more of the mechanical work to AI, and make the underlying data and APIs available to whoever wants to build on top of them. That’s good news for advertisers. It means less time spent stitching together disconnected tools and more time available for strategy.
The advertisers who benefit most won’t be the ones with the fewest tools. They’ll be the ones who know which parts to automate and which parts still need a second set of eyes, especially at enterprise scale, where governance, cross-campaign visibility, and account-level judgment matter most. That kind of portfolio-level view, across a brand’s full retail media footprint rather than one campaign at a time, is where Pacvue has always focused.
Want the full breakdown, plus a chance to ask your own questions? Join Michael Foulkes, Pacvue’s Head of Customer Success, for unBoxed unPacked: Keynote Breakdown & Live AMA on October 7 at 8am PT. Register here.
