Build a no-code ML workflow with Snowflake, Amazon SageMaker Canvas and Amazon Quick – Part 3: Visualizing insights with Amazon Quick Sight

Part 1 covered the Snowflake database implementation setup and established the foundational infrastructure for our no-code machine learning (ML) workflow.

Part 2 walked through the complete data preparation and model building workflow using Amazon SageMaker Canvas, demonstrating how to connect directly to Snowflake data sources, transform and prepare data using Data Wrangler visual transformations, and build a fraud detection model using the XGBoost algorithm.

In Part 3, the workflow comes full circle by integrating SageMaker Canvas predictions with Amazon Quick Sight, now part of Amazon Quick, to create interactive dashboards that combine operational data with ML predictions for fraud detection business intelligence (BI). This post covers how to import Canvas predictions into Amazon Quick Sight as a dataset, build an analysis dashboard, use generative BI capabilities to surface insights through natural language, and publish those insights to stakeholders.

Building dashboards with Amazon Quick Sight

Amazon Quick Sight is a powerful business intelligence service within Amazon Quick that teams can use to build interactive dashboards, perform deep data analysis, and share insights across their organization. Amazon Quick extends these capabilities with generative AI that supports natural language queries over your data, custom AI agents for workflow automation, and collaborative Spaces for organizing files, dashboards, and knowledge bases in one place.

For teams already using Amazon SageMaker Canvas, Amazon Quick provides a direct path from ML predictions to business-ready dashboards without requiring additional infrastructure or custom integrations.

Solution overview

This section walks through the complete workflow for visualizing fraud detection predictions in Amazon Quick Sight, from importing your Amazon SageMaker Canvas predictions as a dataset to building interactive dashboards and leveraging generative BI capabilities for natural language insights.

Prerequisites

Before starting Part 3, complete the following:

Part 1 – Set up a Snowflake account, create the fraud detection database and table, and retrieve the Snowflake organization account name needed for Amazon SageMaker Canvas.

Part 2 – Connect Amazon SageMaker Canvas to Snowflake, prepare the dataset using Data Wrangler, build and train the XGBoost fraud detection model, and send predictions to Amazon Quick Sight.

Sign up for Amazon Quick to facilitate access to Amazon Quick Sight, and upgrade users to Admin Pro, Author Pro, or Reader Pro for generative BI capabilities.

Import Canvas predictions as a dataset

Before building visualizations, import the Canvas predictions into Amazon Quick Sight as a dataset:

In the previous part of this series, Amazon SageMaker Canvas predictions were sent to Amazon Quick Sight. This dataset will be available on Amazon Quick Sight’s console.

From the left navigation pane, choose Datasets to locate your Canvas predictions dataset.

Choose Create analysis. This opens a blank canvas where you can begin building visualizations that transform your fraud detection predictions into actionable insights.

Build visualizations

With the dataset imported, the next step is creating visuals to analyze the predictions:

In the analysis interface, start by selecting fields from your dataset to create your first visual. For detailed guidance on creating dashboards, see the Amazon Quick Sight analysis creation.

The following is the analysis dashboard created showcasing fraud patterns across multiple dimensions including transaction categories, merchant behavior, and temporal patterns.

Figure 1: Fraud detection analysis dashboard showing fraud patterns across transaction categories, merchant behavior, and temporal trends

The generative BI capabilities of Amazon Quick Sight accelerate dashboard creation through natural language interactions.

To access Generative BI capabilities in Amazon Quick Sight, upgrade your account’s users to Admin Pro, Author Pro, or Reader Pro roles within your Amazon Quick subscription. Pro roles grant access to the generative BI capabilities relevant to each assigned role, including the ability to share generative Q&A topics with other users. To understand how subscription names map to user roles, see Understanding Amazon Quick subscriptions and roles.

Choose the sparkle icon. In the pane that opens, the chat displays context-aware options based on your current task. When working in an analysis, you can build calculations, create and edit visuals, set up Q&A topics, or ask questions about your data directly.

Figure 2: Amazon Quick Sight generative BI pane with context-aware options for building calculations, visuals, and Q&A topics

To generate a custom visual, enter a natural language description of the that you want outcome in the Build a visual menu. Review the generated visual and choose Add to analysis to include it in your dashboard.

In the following example, a visual of fraud rate by transaction hour is built using a simple natural language query which reveals how fraudulent activity fluctuates across different times of day, surfacing temporal patterns that would be difficult to detect through manual analysis alone.

Figure 3: Generated visual of fraud rate by transaction hour created from a natural language query

To query the dashboard directly without building a new visual, enter your question in the generative BI pane to get an instant, AI-powered answer. In the following example, the generative BI capabilities responds to a simple query of total fraud cases in the state of Washington and instantly returns the result, demonstrating how natural language questions can surface data-driven answers for business users.

Figure 4: Generative BI answer returning total fraud cases in Washington state from a natural language question

Publish the dashboard

To publish the dashboard, complete the following steps:

After all modifications to the analysis are complete, choose Publish in the top corner of the analysis.

Figure 5: Publishing an analysis in Amazon Quick Sight using the Publish option

In the dialog box that appears, enter a descriptive name for the dashboard in the Dashboard name field. Verify that the Allow executive summary option is selected. This enables AI-powered executive summaries for dashboard viewers. Choose Publish dashboard to complete the publishing process.

Figure 6: Publish dashboard dialog with a dashboard name and the Allow executive summary option selected

Figure 7: From the published dashboard, users can generate an executive summary using the Create option

After the dashboard is published, users can access additional capabilities from the top menu bar:

Share – Control dashboard access and permissions for users and groups.

Send reports – Schedule automated report delivery via email.

Threshold alerts – Configure notifications when metrics exceed defined thresholds.

Export – Download dashboard content as PDF files for offline distribution.

Create executive summary – Generate AI-powered insights from dashboard data.

In addition to these, there are other capabilities available to help users interact with, customize, and share their dashboards, all of which are covered in detail in the Amazon Quick dashboard documentation.

To generate an executive summary, choose Create in the top menu bar and choose Executive summary from the menu.

The executive summary uses generative AI to automatically analyze your visualizations and provide natural language insights, making it more straightforward for stakeholders to quickly understand the most important findings.

Figure 8: AI-generated executive summary providing natural language insights from the dashboard

The fraud detection dashboard brings together Amazon SageMaker Canvas predictions and Amazon Quick Sight to surface fraud patterns across transaction categories, merchant behavior, and temporal trends all in one interactive view. The published dashboard further enables automated report scheduling, threshold-based alerts, and AI-generated executive summaries, empowering both technical and non-technical audiences to make faster, data-driven decisions.

With generative BI, stakeholders can query insights in natural language and access AI-generated executive summaries, closing the loop on the end-to-end no-code ML workflow built across this series.

Conclusion

This three-part series demonstrates how AWS services can work together to take organizations from raw data to actionable business intelligence entirely through visual, no-code tooling.

For organizations managing large volumes of operational data, this workflow removes the traditional barriers between data science and business decision-making. Insights that once required specialized engineering resources are now accessible to analysts, operations teams, and executives through a governed, scalable, and fully managed AWS architecture. Whether the goal is detecting anomalies, forecasting demand, classifying risk, or understanding customer behavior, this approach scales across use cases and industries.

To start transforming your Snowflake data into actionable business intelligence, explore Amazon SageMaker Canvas and Amazon Quick Sight.

References

Part 1 – Build a no-code ML workflow with Snowflake, Amazon SageMaker Canvas and Amazon Quick – Part 1: Setting up your Snowflake environment

Part 2 – Build a no-code ML workflow with Snowflake, Amazon SageMaker Canvas and Amazon Quick – Part 2: Data preparation and model building with Amazon SageMaker Canvas

About the authors