NSF uses an Azure AI-based, agentic solution to streamline the highly regulated audit process for new medicines and therapies. These audits can involve tens of thousands of documents that must be collected, organized, checked for completeness, and summarized against country-specific and international regulations.

The new solution combines the Microsoft Cloud Accelerate Factory approach with several Azure services, including:

• Azure Document Intelligence to scan and verify that all required documents are present.
• Azure OpenAI models and Model Context Protocol (MCP) Servers to classify and sort documents into the correct, regulated folder structures.
• Azure Blob Storage and Azure Cosmos DB to manage structured data and automate version control.

By automating these manual, repetitive steps, NSF has been able to cut audit turnaround time by half or more. Audits that previously took 4–6 weeks can now be completed in about 2 weeks. This means potentially life-changing medications can reach patients and hospitals faster, while NSF’s scientists focus more on higher-value work such as regulatory strategy instead of document handling.

NSF reports that its Azure AI-based tool delivers a “100% truth value” on the content it processes, meaning the summaries and outputs have been factually accurate in their experience. Human experts still review the AI-generated summaries, but their changes are largely cosmetic or style-related rather than corrections to the underlying facts.

The workflow looks like this:

• Azure Document Intelligence extracts and structures text from thousands of documents.
• Azure OpenAI models synthesize that content into cohesive summaries that align with regulatory needs.
• NSF auditors and scientists then review and refine the drafts as needed.

This approach helps NSF minimize the risk of human error that can occur when people manually manage and cross-check very large document sets. At the same time, it keeps subject-matter experts in control of final decisions, ensuring that regulatory judgment and context remain with experienced professionals.

NSF operates in a highly regulated environment with sensitive medical and intellectual property data, so security and compliance are central to its Azure AI design.

Key elements of the security model include:

• Private data tenancy: Data is maintained in a private tenant within SharePoint and then funneled into Azure Blob Storage, keeping information within NSF’s controlled environment.
• Role-based access: Using Microsoft Entra ID and Azure role-based access control (RBAC), NSF limits access to both data and tools strictly to authorized users.
• Closed cloud environment: The entire auditing workflow runs inside the Azure Cloud, with private connections to and within applications to reduce exposure and risk.
• Controlled AI connectivity: Azure Model Context Protocol (MCP) Servers manage how language models interact with external tools and data sources, allowing NSF to decide how networked or isolated each AI solution should be.

Because NSF is a “Microsoft shop,” it also benefits from the interconnected Microsoft 365 ecosystem, including the closed environment of Microsoft 365 Copilot. This setup helps NSF reimagine its audit processes while maintaining the level of trust and data protection expected of an organization that safeguards the “sanctity of the NSF logo.”