Most Amazon agencies run ads through automation software like Perpetua or Pacvue. Rules fire, bids adjust, reports get pulled. But the software only does what it’s told, and someone still has to log in, read the data, decide what it means, and figure out what to do next. That part hasn’t changed. It’s still slow, still inconsistent, and it still scales badly.
We built MazeScale differently. Our workflow is built around Claude (Anthropic’s AI) connected to live Amazon Ads data via a tool called DataDoe. Here’s exactly how it works and why we think it’s a better way to manage Amazon advertising.
The Problem with Traditional Amazon Ads Management
Amazon Ads generates a staggering volume of data. A mid-size account might have hundreds of campaigns, thousands of ad groups, tens of thousands of search terms, and metrics updating daily. Bid decisions, budget pacing, search term harvesting, placement adjustments: every one of these has compounding effects on TACOS.
No human can process all of that data consistently, every single day, without making mistakes or missing things. And yet most agencies try. The result is reactive management: things get fixed when they break, not before.
What DataDoe Does
DataDoe is a data connector that gives Claude direct access to a brand’s Amazon Ads account via their MCP server. It pulls live campaign data (spend, impressions, clicks, conversions, search terms, placement performance, keyword bids, budget pacing) and feeds it to Claude for analysis.
The important thing to understand: DataDoe is read-only. It surfaces the data. Claude analyzes it and produces recommendations. Our team then reviews those recommendations and executes the changes in the account. That’s not a limitation; it’s intentional. Every change is reviewed by a human before it goes live. You get the speed and analytical power of AI with the judgment and accountability of experienced account managers.
How the MCP Server Works
MCP (Model Context Protocol) is the standard that lets AI models like Claude connect to live data sources and tools. Think of it as the pipe between Claude’s brain and your Amazon account. Amazon launched their own Ads MCP Server into open beta earlier this year, which was a significant signal that the industry is moving in this direction. But Amazon’s native server covers the Ads API only and requires multiple API calls to return results, which slows things down. DataDoe’s MCP covers both the Ads API and the full Selling Partner API in a single connection, returns structured results instantly, and handles all the OAuth, token refresh, and rate limiting behind the scenes. Setup takes under five minutes. The result is a faster, cleaner data feed into Claude with broader account visibility from day one. Here is an interesting comparison between the two.
What This Means for Clients
The practical outcome is an account that gets reviewed and actioned every single day, not when someone who is juggling 10+ accounts, gets to it. Issues are caught early. Spend doesn’t drift. Search term lists stay clean. Bids stay calibrated to your actual TACOS target, not whatever they were set at six months ago.
Our Account Managers don’t spend their days pulling reports and reformatting spreadsheets. They spend them reviewing AI-generated analysis, making judgment calls, and working on the things that genuinely require human thinking: strategy, client communication, campaign structure decisions, and reading the market context that no algorithm fully captures.
That’s what AI augmentation actually looks like in a well-run agency. Not replacing account managers. Using AI to make them dramatically more effective, so clients get more rigorous management for less than the cost of a traditional retainer.
The Short Version
DataDoe connects Claude to your live Amazon Ads data. Claude does the analysis, daily, consistently, at a depth no manual process can match. Our team reviews it and executes the right moves. You get clear reporting, a managed TACOS target, and an account that’s actually looked at every single day.
