Normalized claim
Review time: 60% decrease
reduced review time by 60 percent
Flo Health scales medical content review and generation using Amazon Bedrock
Use case typeContent review and revisionUpdated Jul 14, 2026
Flo Health turned an AWS Generative AI Innovation Center proof of concept into a production-grade medical content review and generation system. The system uses specialized AI Judges, RAG grounded in internal and trusted external medical sources, and a multi-stage validation and revision loop to support medical accuracy, legal compliance, and brand style. It reduced review time, increased throughput, and kept human experts in control with cited, traceable outputs.
- Organization
- Flo Health
- Industry
- Healthcare
- Location
- Netherlands
- Published
- July 2026
Reported outcomes
−60%
review timeTime & speed
+200%content throughput−80%routine compliance corrections−70%repeated errors
Strategic outcomes
Scale & capacityExpanded medical content production without adding staffRisk & complianceMaintained medical accuracy and traceable references
Why do we believe this?Outcome claims, sources, and evidence checks
Normalized claim
Content throughput: 200% increase
tripled content throughput without expanding the medical team
Normalized claim
Routine compliance corrections: 80% decrease
reduced routine compliance corrections by 80 percent
Normalized claim
Repeated errors: 70% decrease
reducing repeated errors by over 70 percent
Why do we believe this deployment?Customer identity, provider attribution, maturity, and source checks
- Customer
- Flo Health
- Provider
- AWS
- Maturity
- PoC
- Linked source
- AWS Machine Learning Blog
Flo Health turned an AWS Generative AI Innovation Center proof of concept into a production-grade medical content review and generation system
Customer identity supportedSource describes one deploymentMaturity supported
Primary read
Use case focus
Showing 3 of 3
- 1Content review and revision
- 2Medical writing automation
- 3Workflow orchestration
- Medical experts spent about seven working days per article verifying facts and references.
- Scaling the review team was costly and not sustainable.
- General-purpose AI risked hallucinations and ungrounded medical content.
- Built a production AI review system with specialized AI Judges for different review dimensions.
- Used Amazon Bedrock models including Claude Haiku and Claude Sonnet for different tasks.
- Grounded generation with Amazon Bedrock Knowledge Bases, semantic search, Amazon S3, Amazon API Gateway, and AWS Step Functions.
- Reduced review time by 60 percent.
- Tripled content throughput without expanding the medical team.
- Reduced routine compliance corrections by 80 percent and repeated errors by more than 70 percent.
Architecture
Production system adapted from an AWS Generative AI Innovation Center PoC. Specialized AI Judges evaluate medical accuracy, legal compliance, brand style, and other dimensions. The workflow uses RAG over Amazon S3-hosted knowledge base materials and trusted external medical sources, Amazon API Gateway for real-time UI communication, and AWS Step Functions for orchestration. Claude models in Amazon Bedrock are selected per task (Haiku for classification/routine analysis, Sonnet for higher-fidelity generation and reasoning). Human reviewers remain in the loop with highlighted citations and direct source links.
Sources & evidence1
Evidence: Medium50/100Evidence strength
- Customer explicitly identified
- Deployment status explicitly supported
- Quantified outcome available
- Technical implementation details available
Type: Blog PostPublished: Jul 14, 2026Publisher: AWSEvidence: VendorConfidence: Medium
AI-generated summary. Verify important details with the linked sources before relying on this case.
Explore related AI use cases
Customer evidence
Provider evidence
AI patterns
Browse the catalog
Was this useful?
Community
Comments
No published comments yet.