CardX Source Note

Case Source Notes

CardX Source Note

Back to Articles

This note confirms what is publicly supported and what is not.

What is public

The public CardX article and customer-approved deployment context support these facts:

  • CardX (SCB X Group) publicly describes its hyper-personalization initiative under its "AI with Soul" program
  • the work sits in a regulated financial-services context in Thailand
  • customer relevance, governance, and operational compliance all matter in the workflow
  • the solution is launched and live in production inside CardX's own AWS environment as a Customer-Deployed AWS deployment
  • production launch occurred in October 2025 and the workflow has been continuously in production since
  • the deployment uses Amazon Bedrock, Amazon SageMaker, Amazon EKS, EC2 G6e GPU, S3, AWS Glue, AWS Lambda, Amazon QuickSight, AWS KMS, AWS CloudTrail, and Amazon CloudWatch as its core AWS services
  • the case maps to the Agentic AI Applications competency category — a multi-step orchestrated workflow performing per-customer segmentation, decisioning, tool use, and reviewer routing on AWS, with Generative AI (content-variant generation via Amazon Bedrock) used as one tool inside the agentic chain
  • production status (not pilot, not proof of concept) is confirmed by AWS ACE-registered launched-stage opportunity records, the public CardX article, and customer-approved deployment artifacts

What the case study adds

The case study adds narrow execution detail from customer-approved material:

  • Pixel ML built and operates the modeling, optimization, and content-variant generation layer through its AgenticFlow platform inside CardX's AWS account
  • the system runs in customer-controlled AWS, not vendor SaaS — customer financial PII never leaves the CardX compliance perimeter
  • the work reached production use and is now in continuous operation

What stays private

These details are not on the public site and remain NDA-protected, available only to AWS for validation purposes:

  • exact KPI and conversion uplift figures (profit uplift, take-up rate uplift)
  • pricing, rate/fee economics, and offer-design logic
  • detailed segmentation feature engineering
  • customer-specific architecture and data-pipeline configuration
  • internal compliance and audit-control material

Read next