Summary
AWS and Bluesight have built an AI tool called Prism to help hospitals manage complex 340B drug compliance rules. The tool automates manual data checks that can take thousands of staff hours each year. It is already used by 20 health systems for controlled substance monitoring. A second version for group purchasing organization (GPO) compliance is expected later in 2026.
Main Impact
The new AI layer, called Prism, connects hospital pharmacy and compliance data across Bluesight's product suite. It aims to replace manual work that pharmacy teams do to check if drug purchases follow federal rules. For example, a single hospital can spend over 4,000 staff hours each year reviewing GPO drug purchases. The AI can cut report generation time from six hours to 15 minutes, according to Bluesight's internal tests across 20 health systems.
Key Details
What Happened
Bluesight developed Prism Assistant for ControlCheck, a tool that monitors controlled substances in hospitals. The AI gives users a chat interface to query data, create charts, and generate reports. It was built in a three-day workshop with AWS in September 2025. Eight Bluesight engineers worked with seven AWS specialists to create the first version.
The system uses Amazon Bedrock and Strands Agents. It does not give the AI direct access to hospital databases. Instead, Bluesight wrapped existing ControlCheck APIs in AWS Lambda functions. This keeps business logic safe while the AI interprets questions and gathers results. Query time dropped from five minutes to 10 seconds.
Important Numbers and Facts
- 20 health systems currently use Prism Assistant for ControlCheck.
- Report generation time dropped from 6 hours to 15 minutes (96% reduction).
- Pre-investigation triage fell from 3 hours to 10 minutes.
- Controlled-substance variance analysis dropped from 30 minutes to under 1 minute.
- The GPO agent achieved 100% invoice discovery rate and 93% evidence accuracy in synthetic tests.
- Bluesight reached general availability in under nine months.
Background and Context
The 340B program is a federal rule that helps certain hospitals buy outpatient drugs at lower prices. These hospitals include Disproportionate Share, Children's, and Free-Standing Cancer hospitals. The rule says these hospitals cannot buy drugs through GPO contracts if non-GPO channels can supply the drug. Compliance teams must document exceptions when supply conditions prevent GPO purchases.
Currently, staff must compare purchase data with FDA shortage notices, ASHP records, inventory data, machine learning forecasts, and back-order reports from other hospitals. This is a data-heavy process done manually. The AI aims to automate this by pulling data from Bluesight's products: CostCheck (purchase info), ShortageCheck (drug availability), and 340BCheck (eligibility data).
Public or Industry Reaction
Samir Neyazi, Director of Product Management at Bluesight, said the tool gives diversion program leaders faster answers and removes manual work from investigations. However, experts note that the 100% and 93% accuracy figures come from synthetic tests, not real hospital data. Enterprise buyers should test the system with their own data before relying on it for compliance decisions.
What This Means Going Forward
The AI tool could save hospitals thousands of staff hours each year. But it comes with risks. The system uses a deterministic scoring service for compliance decisions, not the AI itself. This means the AI gathers data and drafts explanations, but a separate service calculates the final compliance score. Hospitals still need to set their own policy rules for things like shortage thresholds and inventory periods.
HIPAA controls are in place. Amazon Bedrock holds HIPAA eligibility, and Bluesight has a Business Associate Agreement with AWS. AWS says it does not train models on customer data. Audit trails record all agent decisions, tool calls, and data access events. This matters when hospitals need to explain why they allowed a GPO purchase.
Hospitals should run historical cases in parallel with existing review processes before using AI results for compliance decisions. Each finding should keep the scoring-rule version, source evidence, and tool trace that produced it.
Final Take
Bluesight and AWS have built a practical AI tool that addresses a real pain point in hospital pharmacy compliance. The 20 health systems already using it show early adoption. But the GPO agent is still in development, and synthetic test results need real-world validation. Hospitals should approach with caution, test thoroughly, and keep human oversight in place.
Frequently Asked Questions
What is the 340B program?
The 340B program is a federal rule that lets certain hospitals buy outpatient drugs at lower prices. It applies to hospitals that serve many low-income patients, children's hospitals, and cancer hospitals. The rule limits how these hospitals can buy drugs through group purchasing organizations (GPOs).
How does the AI tool help hospitals?
The AI tool automates manual data checks that pharmacy teams do to comply with 340B rules. It pulls data from Bluesight's products, uses a chat interface to answer questions, and generates reports. This can cut report generation time from six hours to 15 minutes and reduce pre-investigation triage from three hours to 10 minutes.
Is the AI tool safe for patient data?
Yes, the tool uses HIPAA-eligible Amazon Bedrock. Bluesight has a Business Associate Agreement with AWS. AWS does not train models on customer data. The system uses encryption, authentication, and audit trails to protect data. However, hospitals should still test the tool with their own data before using it for compliance decisions.
