Bluesight, a healthcare compliance platform, is leveraging Amazon Bedrock to develop agentic AI solutions that streamline hospital compliance processes. The company, which serves thousands of partners across the United States, is addressing the growing complexity of drug pricing program compliance by integrating AI across its product suite. This initiative, supported by AWS, aims to reduce the manual effort required for audit processes and improve data analysis for hospital compliance teams. The solution, named Prism, is designed to unify insights from multiple products and automate critical compliance checks. Bluesight’s collaboration with AWS through the EBA program enabled rapid development of a production-ready architecture, allowing the company to deploy its AI solution within nine months. This marks a significant step in the company’s strategy to enhance its offerings through agentic AI capabilities.
Bluesight’s first AI initiative focused on drug diversion detection, where the company sought to automate the analysis of controlled substance transactions. The team aimed to create a conversational interface that could perform this analysis in seconds, significantly reducing the time spent by compliance teams on manual reporting and data correlation. The solution required strict adherence to HIPAA regulations and data security standards, which Amazon Bedrock’s HIPAA eligibility and BAA compliance helped address. Additionally, the use of AgentCore Runtime ensured secure, serverless hosting and session isolation, critical for handling concurrent queries from multiple hospitals. These capabilities allowed Bluesight to integrate live hospital data without the need for custom infrastructure, streamlining the development process.
The collaboration between Bluesight and AWS through the EBA program enabled the company to rapidly prototype and deploy its AI solution. The team, consisting of eight Bluesight engineers and seven AWS professionals, built a functioning agent using Strands Agents on Amazon Bedrock, hosted on AgentCore Runtime. They connected over 10 ControlCheck APIs through AgentCore Gateway, implemented a frontend with chart generation, and added observability for performance monitoring. A key architectural decision was to separate AI reasoning from the data layer, reducing query latency from 5 minutes to 10 seconds. This approach allowed the team to maintain business logic in the application layer while enabling efficient data processing.
Source: awsml
