Learn how to use Claude to diagnose Amazon Ads performance changes, identify anomalies, and uncover what actions to take next.
Stream now
Amazon brands and agencies spend a lot of time trying to understand what changed in their ad accounts, why performance moved, and what action to take next. In this live workshop, Kenton Snyder, Product Manager at Intentwise, showed how AI tools like Claude can help make that analysis faster, more repeatable, and easier to act on.
The session focused on a practical question: What happens when Claude is connected directly to your Amazon data?
Rather than exporting campaign reports, search term reports, product-level data, and Seller or Vendor Central files into large spreadsheets, teams can use AI to ask natural-language questions about performance. When paired with the right data connection and brand context, Claude can help diagnose changes, uncover opportunities, and turn weekly reporting into a more automated workflow.
Kenton started by showing why AI tools need more than raw data to produce useful answers. By default, tools like Claude are built to answer broadly. That means if you ask a generic question, you may get a generic answer.
To make the analysis more relevant, Kenton recommended giving Claude brand-specific context first. That context can include:
In the demo, Kenton created a Claude project for Amazon analysis and uploaded a markdown file with brand context. This allowed Claude to reference the same business background throughout the session, instead of requiring the user to re-explain the account every time.
The session then moved into how Claude can access Amazon data.
Kenton used Intentwise AI Gateway, Intentwise’s MCP solution, to connect Amazon Ads, Seller Central, and Vendor Central data into Claude. With this setup, Claude can query live Amazon data without requiring the user to manually upload CSV files.
Manual uploads are still possible, but Kenton explained that they create more friction. Teams need to download files, upload them into Claude, explain how the files relate to each other, and repeat the process every time they want to run analysis. With an MCP connection, Claude can pull only the data it needs to answer the question.
This is especially helpful when teams want to connect different types of data, such as ad spend, inventory, product performance, campaign data, and total sales.
One of the first examples Kenton walked through was an ACOS diagnosis prompt.
The question was designed to identify which campaigns drove a change in ACOS over the last 14 days, and whether the shift was caused by bids, new keywords, or external factors.
Claude used the connected Amazon data to review campaign-level and search-term-level performance, then surfaced specific patterns behind the change. In the example, Claude identified campaigns that were hurting performance, highlighted changes in clicks and orders, and pointed to possible causes such as increased competitive pressure, new search terms entering broad or auto campaigns, and budget or bid changes.
The value of this workflow is that the analysis does not stop at reporting that ACOS changed. It helps explain what changed, where it happened, and what to investigate next.
Kenton also showed how Claude can help identify opportunities to scale.
In the demo, he asked Claude to find campaigns that had hit their daily budget at least three days in the prior week while keeping ACOS below a target threshold. Claude returned campaigns that appeared to be budget constrained but still efficient, along with recommendations for where budget could potentially be increased.
This is a useful workflow for teams that want to find growth opportunities without manually scanning campaign reports. Instead of looking only for poor performance, Claude can also surface campaigns that are performing well but may be underfunded.
The second half of the session focused on automation.
Kenton showed how the same approach could be turned into a recurring weekly analysis workflow. Using Claude projects, a context file, and an agent-style prompt, teams can define the checks they want Claude to run on a schedule.
Example weekly checks included:
The goal is to move from manually asking questions every week to having a repeatable workflow that delivers the most important findings automatically.
Kenton also showed how the output could be sent into Slack, summarized into a short weekly update, and stored in project memory so Claude can reference past findings over time.
During Q&A, attendees asked how Intentwise AI Gateway differs from the native Amazon Ads MCP.
Kenton explained that one of the biggest differences is data coverage. Amazon Ads MCP is focused on advertising data. Intentwise AI Gateway brings in Amazon Ads, Seller Central, and Vendor Central data together, which makes it possible to connect advertising performance with other business signals.
For example, teams can ask questions that combine ad spend with inventory, total vendor sales, seller sales, or other non-advertising data points. This broader data layer helps Claude answer more complete business questions, not just advertising questions.
The audience also asked about security, especially for European advertisers and teams working with sensitive business data.
Kenton emphasized that security matters on two levels: the MCP connection and the AI tool itself. On the Intentwise side, he noted that the underlying infrastructure is built around enterprise data standards. He also recommended that teams using Claude or similar AI tools make sure they are on plans that prevent model retraining on their data.
For teams with stricter requirements, enterprise plans or contained environments may be necessary. The key recommendation was to avoid putting sensitive business data into AI tools unless the right data protection agreements and settings are in place.
The main takeaway from the session is that Claude can be much more than a place to ask one-off questions. When connected to clean Amazon data and given the right brand context, it can become a repeatable analysis layer for Amazon teams.
Instead of spending hours exporting reports, building pivot tables, and writing up findings, brands and agencies can use AI to ask better questions, diagnose changes faster, and automate parts of their weekly performance review.
For Amazon teams managing large accounts, multiple clients, or complex data across ads, inventory, and sales, this creates a clear opportunity: move from manual reporting to AI-assisted diagnosis, faster insights, and more consistent action.
You don't need to spend hours understanding where—and why—your Amazon performance is changing.
In this session, Intentwise Product Manager Kenton Snyder will showcase how to set up Claude to identify performance anomalies as they happen, what caused them, and what you can do about it.
Register now, and on June 10 at 10 am PST/1 pm EST, we'll show you how to turn Claude into an always-on auditing layer for your Amazon business.
You don't need to spend hours understanding where—and why—your Amazon performance is changing.
In this session, Intentwise Product Manager Kenton Snyder showcases how to set up Claude to identify performance anomalies as they happen, what caused them, and what you can do about it.
We show you how to turn Claude into an always-on auditing layer for your Amazon business.
