Amazon is betting big on AI for advertising. With $17.7 billion in ad revenue in Q3 2025 alone (+24% YoY), the company unveiled a full AI ecosystem at unBoxed 2025 — headlined by the Amazon Ads Agent. Campaign setup dropped from 60+ minutes to under 10. Beta testers saw 18% lower CPM and 16% lower CPA. Here is everything you need to know.
"Crystal Box, Not Black Box"
That is how Amazon VP Kelly MacLean described their approach at unBoxed 2025. Amazon wants AI-powered advertising to be transparent and accessible — not a mysterious algorithm. Five new tools form the ecosystem. Let us break them all down.
1. What Is the Amazon Ads Agent?
The Amazon Ads Agent is an AI-powered assistant built directly into the Amazon Ads console. Announced on November 11, 2025, at unBoxed in Nashville, it represents Amazon's most ambitious push to democratize advertising through artificial intelligence.
Instead of clicking through complex dashboards, setting bid strategies manually, or spending hours analyzing audience segments, advertisers can now simply tell the Ads Agent what they want — in plain English.
How Does It Work?
- Conversational interface — Type natural language commands like "Pause all campaigns with ROAS under 2" or "Create a Sponsored Products campaign for my top 10 ASINs"
- Campaign creation at scale — Upload a media plan and the Ads Agent builds out campaigns automatically
- Smart audience targeting — The AI analyzes thousands of audience segments to find the best match for your products
- AMC Analytics — Converts natural language questions into SQL queries against the Amazon Marketing Cloud, so you can get insights without writing a single line of code
2. The 5 Pillars of Amazon's AI Ads Ecosystem
The Ads Agent does not operate alone. Amazon announced a complete ecosystem of five interconnected AI tools at unBoxed 2025. Together, they cover the entire advertising lifecycle — from creative production to full-funnel optimization to third-party integrations.
Pillar 1: Ads Agent
The conversational AI assistant for campaign management. It sits inside the Amazon Ads console and lets you create, optimize, and analyze campaigns using natural language. Think of it as your always-available advertising co-pilot.
Pillar 2: Creative Agent (Open Beta)
The Creative Agent generates display ads, video ads, and audio ads directly from your product pages. It has access to 50+ creative tools, including a video generator. What used to take weeks and thousands of dollars with an agency can now be done in hours — at no extra cost.
Professional Creatives Without the Agency Bill
The Creative Agent produces display, video, and audio ads from your product listing. It expanded to Canada and the UK in February 2026. No design skills required, no additional fees.
Pillar 3: Campaign Manager (Global Since Dec 2025)
Campaign Manager unifies the Amazon Ads Console and DSP into a single interface. You can manage up to 20 accounts simultaneously. Its standout feature: Guidance Cards — click-to-apply AI recommendations that suggest optimizations and let you implement them with a single click.
Pillar 4: Full-Funnel Campaigns (Open Beta, Q1 2026)
This is where it gets interesting for sellers. With Full-Funnel Campaigns, you provide one prompt — your business goals, budget, and campaign duration — and the AI recommends optimal budget distribution across Sponsored Products, Sponsored Brands, Display, and Streaming TV. One brief, full-funnel coverage.
Pillar 5: MCP Server (Open Beta, Feb 2026)
The Model Context Protocol (MCP) Server is the open interface. Developed on the MCP standard by Anthropic, it acts as middleware between external AI agents and the Amazon Ads API. It works with Claude, ChatGPT, Gemini, and other AI models, providing 50+ tools for campaign management.
3. What Can the Ads Agent Actually Do?
Let us get specific. Here is a comparison of manual campaign management versus the AI-assisted approach:
| Task | Manual | With Ads Agent |
|---|---|---|
| Campaign setup | 60+ minutes per campaign | Under 10 minutes |
| Audience targeting | Manual segment research | AI analyzes thousands of segments |
| Campaign insights | Hours of data analysis | 67% faster via natural language |
| AMC queries | Write SQL manually | Natural language to SQL |
| Bulk optimization | Edit campaigns one by one | "Pause all campaigns with ROAS under 2" |
| Creative production | Weeks + agency fees | Hours, no extra cost |
| Full-funnel allocation | Manual budget splitting | AI-optimized distribution |
| Multi-account management | Switch between consoles | Up to 20 accounts in one view |
The numbers from the beta are promising. However, it is important to note that these are averages from early adopters — results will vary depending on category, competition, and how well you use the tools.
4. The MCP Server — The Open Interface
The MCP Server deserves special attention because it signals a fundamental shift in how Amazon approaches third-party integrations.
The Model Context Protocol was developed by Anthropic as an open standard for connecting AI models to external tools and data sources. Amazon adopted it as the foundation for their Ads API integration layer.
What Does the MCP Server Do?
- Middleware layer — It sits between any AI agent (Claude, ChatGPT, Gemini, custom agents) and the Amazon Ads API
- 50+ tools — Campaign creation, bid management, reporting, audience targeting, and more
- Standardized interface — Any MCP-compatible AI agent can interact with Amazon Ads without custom API integration
- Open beta since February 2026 — Available to advertisers and developers globally
Technical Architecture
The MCP Server translates natural language intents into structured API calls. When you tell your AI agent "Show me my top 5 campaigns by ROAS this month," the MCP Server converts that into the appropriate Amazon Ads API queries, retrieves the data, and returns it in a format the AI can interpret and present to you.
5. New AI Agent Policies for Sellers (March 2026)
The new AI Agent Policy, effective March 4, 2026, establishes three core requirements:
- Self-identification — All AI agents must identify themselves as automated systems when interacting with Amazon Ads. No pretending to be human operators.
- Policy compliance — AI agents must comply with all existing Amazon Advertising policies, including content guidelines, targeting restrictions, and spending limits.
- Access revocation — Amazon reserves the right to revoke access for any AI agent that violates these policies. This includes both temporary suspension and permanent bans.
Double Standard?
It is worth noting that Amazon's own AI tools — the Ads Agent, Creative Agent, and Campaign Manager — operate freely within the platform, while external AI agents face strict identification and compliance requirements. Amazon controls the playing field and plays on it at the same time. Sellers should be aware of this asymmetry when building their AI advertising stack.
6. Opportunities and Risks
Opportunities
- Democratization — Small and mid-size sellers get access to enterprise-level tools at no extra cost
- Time savings — Campaign setup reduced by 80%+, insights generated 67% faster
- Cost reduction — 18% lower CPM and 16% lower CPA in beta testing, plus free creative production
- Full-funnel access — Streaming TV ads and DSP campaigns are no longer reserved for big brands with big budgets
- Third-party flexibility — The MCP Server allows you to build custom AI workflows tailored to your business
Risks
- Natural language misinterpretation — "Pause all campaigns with ROAS under 2" sounds simple, but what if the AI misinterprets your timeframe, currency, or campaign scope? A single ambiguous command could pause your entire ad portfolio
- Over-reliance on AI — Gartner warns that 40%+ of agentic AI projects will be abandoned or scaled back by end of 2027 due to unmet expectations. The technology is powerful but not infallible
- Competitive compression — When everyone has access to the same AI tools, the competitive advantage shifts from tool access to strategy and product quality. AI levels the playing field — which means it gets harder to differentiate through advertising alone
- Platform dependency — Amazon's AI tools work only on Amazon. Sellers building their entire advertising strategy around these tools increase their dependency on a single platform
- Policy risk — Third-party AI agents face access revocation risks under the new policy. Building workflows on external tools means accepting this uncertainty
7. Timeline — All Key Dates
First preview of the AI creative tool for generating display, video, and audio ads from product pages.
Amazon unveils the complete AI advertising ecosystem in Nashville: Ads Agent, Creative Agent, Campaign Manager, Full-Funnel Campaigns, and MCP Server.
Unified Ads Console + DSP interface goes live worldwide. Guidance Cards with click-to-apply recommendations included.
One prompt, full-funnel coverage. AI distributes budget across SP, SB, Display, and Streaming TV.
The Model Context Protocol Server enters open beta, enabling Claude, ChatGPT, Gemini, and other AI agents to manage Amazon Ads.
The Creative Agent becomes available in Canada and the United Kingdom.
Self-identification, policy compliance, and access revocation rules go into effect for all AI agents interacting with Amazon Ads.
8. What Sellers Should Do Now
Amazon's AI advertising ecosystem is live and expanding. Here is your action plan:
- Try the Ads Agent now — Log into your Amazon Ads console and start with simple commands. Create a test campaign using natural language to see how it works in your category.
- Test the Creative Agent — Generate display and video ads from your best-performing product pages. Compare the AI creatives to your existing ads in A/B tests.
- Explore Full-Funnel Campaigns — If you have only been running Sponsored Products, use Full-Funnel Campaigns to test Sponsored Brands and Display without manually splitting budgets.
- Review the new AI policy — If you use any third-party AI tools for advertising, ensure they comply with Amazon's March 2026 requirements.
- Evaluate the MCP Server — If you work with a developer or use AI assistants for business operations, the MCP Server could automate significant parts of your advertising workflow.
- Do not automate everything at once — Start with one campaign type, measure for 2-4 weeks, then expand. Gartner's warning about abandoned AI projects should keep you grounded.
The sellers who will benefit most are those who adopt early but adopt wisely — testing, measuring, and iterating rather than handing over their entire ad budget to an AI on day one.
Optimise your Amazon ads
Sellercore gives Amazon sellers the tools to grow: repricing, ads optimisation, inventory and profit analytics in one place. Try it free today.
Get started free