In the span of three days, Amazon, Meta, Discord and Google each described the same change to how advertisers hand work to a platform: the advertiser states a goal, and software handles more of the steps between the goal and the live campaign.

The announcements came from different companies with different products. Amazon Ads published its recap of the unBoxed 2026 event in San Francisco on September 30. Meta announced Muse for Small Business on September 29. Discord launched Ads Manager and Video Quests on September 29. Google introduced Vertical Video Unification in Display & Video 360 on September 28. Read one at a time, each is a product update. Read together, they show where the work of media buying is moving and where the human sign-off is being placed.

What the four announcements have in common

Each launch removes a setup step that used to belong to the advertiser. Amazon is folding the tools that were separate into one buying experience. Meta’s agent takes a goal and does tasks across connected business tools. Discord’s new platform takes an objective and bids toward it. Google’s platform buys vertical video from many publishers in one workstream, with Gemini running buying, execution and measurement.

Advertisement

MarTech Your brand belongs here. Reach the decision-makers who read MarTech every day. Premium placements across the site and newsletter. Advertise with us

The four also differ in an important way, and the differences are the useful part. They put the human control point in different places: at the approval step, in the choice of lane, in the bid, and, at Google, in a place the announcement does not specify. A marketing team choosing among them is choosing where it wants to keep its hands on the wheel.

Amazon: one agent where there were two consoles

The most complete version is Amazon’s. In its recap of unBoxed 2026, Amazon Ads described Amazon Ads Agent as “the unified buying experience that brings together what was formerly the DSP and Ads Console.” The recap lists Sponsored ads, Full-Funnel Campaigns and DVA+ as the products inside it. We covered the console merger when it was first reported in Amazon Folds Its DSP and Ad Console Into One Agent.

Paul Kotas, SVP of Amazon Ads, said: “Amazon Ads Agent understands your end-to-end business, and all your advertising activity.” He added that advertisers can “ask questions, gain insights, consider detailed recommendations, or simply give instructions.” The conversational layer is in an expanded closed beta, according to the recap.

Full-Funnel Campaigns is where the goal-first model is most explicit. Amazon’s description reads: “You set your goal and AI handles the channel mix, creative, audiences, and optimization.” Amazon reported that beta testers saw 67% higher long-term return on ad spend and 29% lower cost to acquire new-to-brand customers. Those are Amazon’s figures from its own beta, and the recap does not describe how the comparison group was built, so they are a vendor claim and not an independent result.

DVA+ applies the same idea to media types. Kelly MacLean, VP of Engineering, Science, and Product, described it this way: “One objective, one budget, one campaign across all media types.” She also addressed the control question directly: “Some of you want full control over every decision. Others want us to optimize for you. We now make it easy to do both.” The recap says DVA+ offers a streamlined path and advanced settings that expose the programmatic levers.

Two more items in the recap point at the agent-facing side. Amazon Ads MCP Server (Lite) starts with three core tools that help agents discover and invoke other capabilities without loading the full toolset up front, and an AMC Expert Tool interprets natural-language requests for analytics work. Both are aimed at software that calls Amazon Ads on an advertiser’s behalf, not at a person clicking through a console.

Meta: the goal comes first and the approval gate stays

Meta’s September 29 post introduces Muse for Small Business, an expansion of the Muse agent that Meta introduced earlier in September in the US and Canada. The post says: “Give Muse a goal” such as running a business or finding new customers, and the agent gets the work done. It connects to tools including Asana, Box, Canva, Dropbox, Figma, Granola, HighLevel, Intuit QuickBooks, Klaviyo, Lovable, Notion, Shopify, Slack, Stripe and Zoom, along with Facebook and Instagram business accounts. Our earlier brief on the Klaviyo and Shopify connections is at Meta’s Muse Agent Now Connects to Klaviyo and Shopify.

Meta places the human control point at approval. Its post states: “You’re in control: nothing publishes, sends, or spends without your approval.” That is a specific design choice. The agent can draft, analyze and prepare, and the person authorizes anything that leaves the building or costs money.

Cameron Adams, co-founder and CPO of Canva, said in the post: “The real magic happens when it’s fully connected to your existing work, your context and your brand.” The quote also explains why the connector list matters. An agent with a goal and no context produces generic output. An agent connected to a storefront, a books system and an email platform has the material to produce something specific.

Meta says Muse is free for most of what people need, with subscription plans for those who want more. It lists use cases that include building a growth plan from sales, campaign and social data, and improving ads and content.

Discord: the goal becomes a bid

Discord’s press release from September 29 describes Discord Ads Manager as the company’s first campaign management platform and a self-serve one, covering “forecasting to building a campaign with objective-based bidding to monitoring the campaign’s performance through reporting.” Delivery is auction-based with a dynamic cost model. In Discord’s words: “Rather than a fixed CPM, ads compete in an auction with pricing that fluctuates dynamically.”

Access is staged. Ads Manager is available now with managed access for a select group of advertisers, and full self-serve access is planned for early 2027. Discord says self-service extends advertising to smaller studios and marketers that previously needed a direct sales relationship.

Video Quests, the mobile-first rewarded video format, sits inside Ads Manager. Players watch for approximately 15 seconds and earn Orbs. Discord says the format gives advertisers “a stronger signal of intent based on how long the player chooses to watch.” It also reports that Quests have delivered a 97% median ad completion rate, citing Discord internal data dated June 30, 2026. The original Video Quests product is now evolving into Video Quests Reserve, a reservation-based offering with delivery and rewards tied to full video completion.

Nabib El-Rahman, VP of Product, Business at Discord, framed the aim: “We want advertising on Discord to work for a studio launching its first game just as well as it does for the world’s biggest advertisers.”

The control point here is the bid. An advertiser picks an objective and adjusts bids in real time while watching down-funnel metrics, and the press release says reporting draws on first and third party data. Discord says it has more than 90 million daily active users as of Q2 2026.

Google: the default does the buying, the settings hold the control

Google’s September 28 post, written by Marta Martinez, Managing Director, Google Marketing Platforms, introduces Vertical Video Unification in Display & Video 360. The post says it makes it possible “to buy and measure vertical video programmatically across channels in one platform” and that it centralizes vertical video buys from publishers into a single workstream. Gemini powers buying, execution and full-funnel measurement.

Google reports that during initial testing, Unilever’s Hellmann’s and WPP Media achieved a 24% increase in unique reach while reducing cost per unique user by 25%. The post presents these as results from early testing. Our brief on the launch is at Google Folds Vertical Video Buying Into DV360.

Google’s post centers on unification more than on a conversational interface. Its stated benefit is delivering competitive campaign performance “without overexposing users to the same ads,” which is a claim about frequency across sellers. The post does not describe where an advertiser overrides the system, so that question is one to put to a Google representative before committing budget.

Where the human sits in each design

This section is our reading, not the companies’ claim. Across the four launches the human role falls into four positions.

At Meta, the person approves. Nothing publishes, sends or spends until they say so. That gives the strongest guarantee about what leaves the account and the weakest about speed, since every action waits on a person.

Newsletter

Get the week's best tech coverage.

Free. Read by thousands of HR, tech, and business leaders.

At Amazon, the person chooses a lane. DVA+ offers a streamlined path and advanced controls, and MacLean’s remark that the company makes it “easy to do both” is a statement about choosing how much to delegate. The delegation level is a setting per campaign.

At Discord, the person sets the objective and manages the bid. The auction means price is a live variable, so the human work is watching down-funnel metrics and adjusting.

At Google, the post leaves the human role undefined. It says Gemini powers buying, execution and full-funnel measurement, and the override points are not described.

These positions are not interchangeable. A team that must show an auditor who approved a spend wants Meta’s model. A team that runs many campaigns across media types wants Amazon’s lane choice. A team that buys on performance wants Discord’s bid control. A team with a frequency problem across vertical video sellers will look at Google’s unification. The same company may want different positions for different budgets.

What it means for the marketing leader

The work that used to sit in campaign setup does not disappear. It moves to places that are less visible in a console.

Goals become the input, so goal quality becomes the constraint. Amazon’s Full-Funnel Campaigns asks the advertiser to set a goal and leaves channel mix, creative, audiences and optimization to AI. If the goal is “more sales” and the business actually cares about new customers or margin, the system will optimize toward the wrong thing without any error message. Writing goals that match the P&L is now a skill on the critical path.

Approval rights need an owner. Meta’s design assumes someone reviews each action. In a team of ten, that person needs a defined queue, a turnaround time and authority to reject. We argued the same point from the operations side in Own the Approval Step Before Agents Own Your Stack.

Vendor-reported lifts need your own test. The 67% ROAS figure, the 29% cost reduction, the 24% reach increase and the 25% lower cost per unique user all come from the companies selling the product. Each is useful as a hypothesis. None should set a budget until a holdout in your own account reproduces a result.

Measurement has to span the seams. If buying consolidates inside one platform, so does the platform’s own reporting. Discord’s press release mentions first and third party reporting, and Google’s post ties buying to full-funnel measurement. Keep an independent measurement path so the buyer and the grader are not the same system.

What to do this quarter

Start with an inventory. List which of your paid channels now offer a goal-based or agent-driven path, and which of those you have turned on. For Amazon, the conversational layer is in expanded closed beta, so access may need a request. For Discord, full self-serve access is planned for early 2027, so a managed-access conversation is the only route today.

Then pick one campaign per platform where a mistake is cheap, and run the goal-based path against your current setup with a fixed budget and an identical flight window. Write down the goal text you gave the system, because the goal is the variable you will want to revisit.

Decide who holds approval for each agent. Name a person, define what they can reject, and set a response time. Where a platform lets you require approval before spend, as Meta says Muse does, turn that on first and relax it later with evidence.

Finally, set a reporting rule before launch: any vendor-reported lift gets checked against a holdout or a pre-agreed baseline in your own data. The companies in this piece reported lifts in their own announcements. The advertisers who benefit will be the ones whose own numbers agree.

Source: Amazon Ads, The 11 biggest takeaways from unBoxed 2026. Also drawn from Meta, Muse for Small Business, Discord, Ads Manager and Video Quests and Google Marketing Platform, Vertical Video Unification.