Get More Out of AMC Using AI Agents
Continue learning: Access the webinar resource hub for the white paper, practical tools, and additional resources to help you apply the ideas covered in this session.
Link: Explore the webinar resources
Introduction
Amazon Marketing Cloud (AMC) gives advertisers a deeper view of how shoppers interact with their brands across advertising, organic activity, and different stages of the purchasing journey. The challenge is turning that complex data into insights teams can understand and act on.
In this session, Intentwise’s Kenton Snyder joined host Rolando Galeana to demonstrate how AI agents can make AMC analysis more accessible. Kenton explored how teams can use natural-language questions to analyze multiple datasets, identify patterns, create visualizations, and move from raw data to practical advertising decisions.
The opportunity is not simply to access data faster. It is to ask better business questions, uncover relationships that standard reports may miss, and make AMC insights easier to apply across advertising and analytics workflows.
What you’ll learn
- Why standard attribution can provide an incomplete view of the customer journey
- How AI agents can make complex AMC analysis more accessible
- Which AMC datasets can be analyzed together to answer deeper questions
- How advertisers can translate AI-assisted analysis into media decisions
- Why data quality, precise prompts, and human validation still matter
AMC shows more of the customer journey
Standard advertising reports typically evaluate performance within separate channels and assign conversions based on a particular attribution model. While this makes campaign reporting easier to understand, it can obscure how different advertising interactions contributed to a purchase.
For example, a shopper might first encounter a Sponsored Brands ad, later receive a display ad, search for the brand, and finally purchase after clicking a Sponsored Products ad. A last-touch attribution model may credit the Sponsored Products interaction, even though earlier ads helped introduce and move the shopper toward the purchase.
AMC is a privacy-safe clean room that allows advertisers to analyze pseudonymized advertising and shopping events across the customer journey. Depending on the datasets available, advertisers can evaluate activity such as:
- Advertising impressions and clicks
- Product detail page views
- Add-to-cart activity
- Sponsored ads and Amazon DSP exposure
- Purchases and repeat purchases
- New-to-brand behavior
- Paid and organic interactions
This broader view allows teams to investigate questions that are difficult to answer through standard campaign reports alone. Which ad types work best together? Which interactions introduce new customers? How do different paths affect purchase rates, average order value, or long-term customer value?
AMC provides the data foundation for answering these questions, but accessing and interpreting those insights has traditionally required considerable technical and analytical work.
AI agents can reduce the analytical burden
AMC analysis often involves SQL-based queries, carefully defined date ranges, multiple reports, and large data tables. Prebuilt query libraries can make the process easier, but teams still need to run the reports, combine the outputs, and determine what the results mean.
AI agents can help simplify parts of this workflow. When an AI tool has access to approved, properly structured datasets, users can ask questions in natural language instead of manually reviewing every table or building each analysis from scratch.
During the session, Kenton demonstrated how an AI agent could help users:
- Compare new-to-brand performance across different periods
- Identify products with meaningful changes in customer acquisition
- Analyze path-to-conversion and ad-overlap reports together
- Compare first-touch and last-touch attribution
- Visualize lifetime value and new-to-brand performance
- Examine how purchase behavior changes at different advertising frequencies
The AI agent can help retrieve relevant information, combine datasets, summarize patterns, and present findings in a format that is easier to understand.
This does not eliminate the need for AMC expertise. AI agents still depend on accurate data, clear definitions, and sufficient context. Their conclusions should also be reviewed by someone who understands the business question and the limitations of the underlying reports.
Combining AMC datasets can uncover deeper insights
One of the most valuable applications of AI-assisted AMC analysis is the ability to examine related datasets together.
Path to conversion and ad-type overlap
A path-to-conversion report shows the sequence of advertising interactions that occurred before a purchase. An ad-overlap report evaluates how performance changes when shoppers encounter multiple ad types, regardless of the order.
Combining these perspectives can help advertisers identify both the most effective conversion paths and the ad combinations contributing to those results.
In Kenton’s sandbox demonstration, shoppers exposed to both sponsored ads and DSP performed better than groups exposed to either channel independently. The example suggested that DSP may be more effective when supporting a broader advertising journey than when it is evaluated in isolation.
Because the demonstration used sandbox data, the results should not be treated as a universal benchmark. Advertisers should perform the analysis using their own data before adjusting their strategy.
New-to-brand acquisition and lifetime value
A product that attracts many first-time customers is not automatically the most valuable acquisition vehicle. Teams must also consider whether those customers generate meaningful value after the initial purchase.
Analyzing new-to-brand performance alongside customer lifetime value can help identify products that introduce customers to the brand while also attracting higher-value shoppers.
This analysis can help answer questions such as:
- Which products are the strongest entry points into the brand?
- Are high-acquisition products also generating valuable repeat customers?
- Which products may justify additional advertising investment?
- Where is the brand acquiring volume without sufficient long-term value?
First-touch and last-touch attribution
Comparing first-touch and last-touch attribution can show how different ad types contribute at different stages of the customer journey.
In the session’s demonstration, Sponsored Brands performed more strongly as an initial interaction, suggesting a role in customer discovery and prospecting. Sponsored Display performed better as a final interaction, indicating a potentially stronger role in retargeting.
This type of analysis helps advertisers evaluate a channel based on the role it performs instead of judging every campaign exclusively by last-touch return on ad spend.
DSP frequency and audience performance
AMC can also help teams examine how purchase rates, acquisition costs, and other results change as advertising frequency increases.
The objective is not to adopt a universal frequency cap. It is to identify the exposure range where a specific brand balances reach, conversion, and cost most effectively.
An AI agent can visualize these relationships and highlight where additional impressions may begin producing diminishing returns. Advertisers can then use those findings to evaluate frequency caps, exclusions, bid adjustments, or audience strategies.
Better questions create better analysis
AI agents can make AMC analysis more accessible, but easier access does not automatically produce better decisions.
The quality of the output depends heavily on the quality of the question. A vague request such as “Analyze my AMC performance” gives the AI agent little direction. A stronger prompt identifies the timeframe, audience, comparison groups, metrics, and business decision involved.
For example:
For June 2026, compare shoppers exposed to sponsored ads only, DSP only, and both channels. Evaluate reach, purchase rate, return on ad spend, and new-to-brand rate for each group.
This prompt gives the agent a defined population, comparison, timeframe, and set of success metrics.
Teams also need to understand what the available data can and cannot answer. AMC data is specific to the advertiser’s brand and does not expose another brand’s customer-level activity. Audience creation also remains subject to Amazon’s privacy requirements and minimum audience thresholds.
As Kenton explained during the session, an AI workflow may analyze AMC alongside other approved datasets when those sources are properly connected. However, teams should clearly distinguish between information originating in AMC and information coming from another source.
How to put these ideas into practice
- Begin with a business decision. Define what you are trying to improve, such as budget allocation, audience strategy, customer acquisition, or campaign frequency.
- Identify the required datasets. Determine whether the question requires path-to-conversion, ad-overlap, new-to-brand, lifetime-value, DSP, organic, or other approved data.
- Write a specific prompt. Include the timeframe, products or audiences, comparison groups, metrics, and desired output.
- Request analysis and visualization. Ask the AI agent to explain the most meaningful differences and present them in a format stakeholders can quickly understand.
- Validate the findings before acting. Confirm that the correct data, definitions, date ranges, and attribution models were used. Treat the AI agent as analytical support, not as an unquestionable source of truth.
Key takeaways
- Standard attribution can miss earlier interactions that contributed to a purchase.
- AMC provides a broader view of how advertising and shopping events interact.
- AI agents can make it easier to explore AMC data and analyze multiple reports together.
- Combining datasets often produces more useful insights than evaluating reports separately.
- Results from a sandbox or another advertiser should not be treated as universal benchmarks.
- Precise questions, accurate data, and human validation remain essential.
