NSF accelerates regulated medical audits using agentic AI with Azure Document Intelligence and Azure OpenAI
NSF, a nonprofit scientific and regulatory auditing organization, needed to organize, verify, summarize, and synthesize tens of thousands of documents across country-specific rules for medical audits. The solution uses Azure Document Intelligence, Azure OpenAI, Azure Model Context Protocol (MCP) tools and servers, Azure Blob Storage, Azure Python SDK, Azure Cosmos DB, Microsoft Entra ID, and Azure RBAC to automate document validation, structured sorting, version tracking, and summary drafting. Staff review and refine the AI-generated summaries, while the workflow remains inside Azure cloud controls and private tenant security. The implementation reduced audit turnaround time from 4–6 weeks to about 2 weeks and was reported to deliver near-perfect accuracy for the proof of concept.
- Organization
- NSF
- Industry
- Healthcare
- Location
- United States
- Published
- June 2026
Reported outcomes
−50%
audit turnaround timeTime & speed
Strategic outcomes
Primary read
Use case focus
Showing 3 of 4
- 1Document Processing
- 2Audit Automation
- 3Workflow Automation
- Highly regulated audits needed tens of thousands of documents organized, verified, summarized, and synthesized.
- Manual work created high time burden and risk of human error.
- NSF wanted to expedite audits while preserving compliance and security.
- Built an agentic Azure AI workflow with Azure Document Intelligence to verify required document components.
- Used Azure OpenAI and MCP tools to sort documents into a regulated structure and generate summary drafts.
- Used Azure Blob Storage, Azure Cosmos DB, Azure Python SDK, Microsoft Entra ID, and Azure RBAC for storage, version tracking, and access control.
- The Microsoft Cloud Accelerate Factory helped deliver a proof of concept in 12 weeks.
- Halved or better audit turnaround time from 4–6 weeks to around 2 weeks.
- The tool reportedly delivered 100% truth value with only cosmetic/style edits.
- Reduced risk of human error and enabled plans to scale the workflow to other audit types.
Architecture
NSF built an agentic Azure AI workflow that ingests documents from a private SharePoint tenant into Azure Blob Storage, uses Azure Document Intelligence to verify required components, uses Azure OpenAI and Azure Model Context Protocol (MCP) tools to sort data into a regulated structure, uses Azure Python SDK and Azure Cosmos DB to automate document version tracking, and uses Azure Document Intelligence plus Azure OpenAI to draft audit summaries for human review. The workflow is secured with Microsoft Entra ID and Azure RBAC and runs within the Azure cloud.
Implementation partners1
Sources & evidence1
AI-generated summary. Verify important details with the linked sources before relying on this case.
Explore related AI use cases
Was this useful?