60% faster insights and 80% less manual reporting with embedded generative-AI analytics

Client story

We implemented a scalable, AI-driven business intelligence platform for a SaaS client by integrating Amazon Redshift with Amazon QuickSight Q. The platform delivers advanced analytics, multi-tenancy and generative AI, with the BI experience embedded directly in the client’s application.

Challenge

The client, a multi-tenant SaaS platform, faced:

• Siloed data across tenants with no unified analytics
• Manual reporting processes and limited self-service BI
• Security concerns around data isolation and user access
• Demand for AI-powered insights to support decision-making

Solution

We designed a serverless BI architecture built on:

• Amazon Redshift for centralized data warehousing
• Amazon QuickSight Q for generative-AI analytics
• Keycloak for SSO and identity management
• Row-level security (RLS) for multi-tenant data isolation
• Embedded dashboards and Q&A inside the client’s application

Implementation ran in five steps:

1. Redshift and QuickSight integration
   • Migrated raw data from disparate sources to Amazon Redshift using AWS DMS
   • Created materialized views and automated refresh schedules to unify data for BI workloads
   • Connected Redshift datasets to QuickSight for real-time analytics
2. Keycloak SSO configuration
   • Federated QuickSight access with Keycloak for single sign-on for enterprise users
   • Mapped Keycloak user groups to IAM roles for secure, role-based access to dashboards
3. Multi-tenant security with RLS
   • Implemented row-level security to isolate tenant data within shared datasets
   • Configured tenant-specific filters using Redshift views and QuickSight RLS rules
4. Embedded dashboards and console
   • Used the QuickSight JavaScript SDK to embed interactive dashboards in the front end
   • Published tenant-specific dashboards with dynamic URL parameters for personalized views
5. Generative Q&A
   • Embedded Amazon Q in QuickSight so users can ask questions in natural language, for example “Show top-selling products by region”
   • Enabled multi-visual Q&A so users explore data through AI-generated insights and executive summaries

Architecture and tech stack

Source systems replicate into Redshift through AWS DMS, with AWS Glue and materialized views preparing datasets for QuickSight. Keycloak federates identity and maps groups to IAM roles; RLS rules in Redshift views and QuickSight keep tenants apart in shared datasets; and the QuickSight JavaScript SDK embeds dashboards and Amazon Q in the client’s application.

• Amazon Redshift
• Amazon QuickSight Q
• AWS Glue
• AWS DMS
• Keycloak
• AWS IAM
• QuickSight JavaScript SDK

Results

• 60% faster insights with AI-driven visualizations and automated reporting
• Secure multi-tenancy via RLS, with data isolation and no duplicated infrastructure
• 80% less manual effort through self-service Q&A and embedded dashboards
• A unified user experience with SSO and customized analytics inside the application

Combining Redshift’s scalability with QuickSight Q’s generative AI turned the client’s BI into a modern, secure and user-centric platform. Data access is now open to every tenant, operational overhead is lower, and the client is positioned to apply AI to predictive analytics next.

What made it work

Erbis’s takeaways from this engagement:

• Isolate tenants in the data layer, not in copies of it. Row-level security in shared datasets delivered separation without multiplying infrastructure.
• Make identity the backbone before embedding anything. Mapping Keycloak groups to IAM roles is what made embedded dashboards safe to expose.
• Embed analytics where the work happens. Dashboards and Q&A inside the application, not in a separate BI tool, are what drove the reduction in manual reporting.
• Start natural-language access on a governed warehouse. Generative Q&A is only as trustworthy as the materialized views and refresh schedules beneath it.