NSF enables life-saving treatments to get to patients faster with Azure AI

When you pick up a prescription—perhaps a medication to manage a chronic illness or one to ease a loved one’s pain—you trust that the treatment is both safe and will do what it promises. This confidence comes from years, even decades, of rigorous testing and verification long before the medication ever reaches your hands. The nonprofit NSF is an independent, science-based organization that enables pharmaceutical companies, as well as many other industry businesses, to undergo exacting review to meet regulatory requirements and bring vital products to consumers.

“For more than 80 years, we have been at the forefront of the global effort to improve human health,” says Heather Wilberger, VP and Chief Information and Digital Officer at NSF. The organization reviews and audits products, claims, and practices worldwide, from medicine and water filters to organic designations and biomedical equipment.

NSF saw an opportunity to use AI to streamline time-intensive, manual, and repetitive tasks to free scientists and other professionals to expedite audits and allow more time for higher level work. 
After several attempts to develop an AI agentic tool on their own, NSF turned to the Microsoft Cloud Accelerate Factory, the no-cost service that employs experts to kickstart Azure projects. While NSF had budgeted a year to develop an AI solution to streamline audits in its life science division, Microsoft developers delivered a successful proof of concept in just 12 weeks.  

“We look at this project not only from an auditing, regulatory, and review perspective, but as an opportunity to grow,” explains Liam Rogers, Global Associate Director of Innovation Technology at NSF. “The quicker we can turn around work, the more of an impact we can have across the global community, the more positive outcomes we can generate.”

Each faster audit translates to a potentially life-saving or game-changing medication reaching patients faster, explains Kush Dhody, President of Amarex Clinical Research, a company within NSF. He adds, “Each product we enable to be brought to market could affect the lives of millions of people. That’s what drives the work we do.” 

Increasing efficiency to get medicine to market

New products, medicines, and therapies go through stringent review processes at multiple checkpoints to ensure they are safe and effective. NSF audits products on behalf of clients to help them navigate this regulatory maze, lending confidence to all stakeholders that the products perform at every stage.  

In other words, “We help pharmaceutical, biotech, and medical device companies bring safer and effective products to market so patients who need them can access them,” Dhody says. “Through Microsoft products and this AI tool, we are expediting our regulatory processes to help our clients—and by extension patients—benefit from these therapies faster.”  

Audits often involve tens of thousands of documents, which must be accounted for, organized, reviewed, summarized, and synthesized in alignment with a wide variety of country-specific and international regulations. By meticulously organizing the content, enacting version control, ensuring compliance, and drafting summaries, the Azure AI-based agentic solution is reducing the department’s audit turnaround time by half—or more. While the average audit used to take four to six weeks, staff using the AI solution now need only two weeks to complete this process.  

The agentic solution draws on structured data collected in Azure Blob Storage. First, Azure Document Intelligence combs through the many documents to verify NSF has every needed piece. Then Azure OpenAI models and Model Context Protocol (MCP) tools within the AI solution sort the data and documents into the correct folders within an internationally regulated structure. Meanwhile, a combination of Azure Python SDK, Azure Document Intelligence, and Azure Cosmos DB automates version tracking of each document. Finally, Azure Document Intelligence scans through the documents and sends raw text to Azure OpenAI models, which generate summary drafts. This step synthesizes conclusions from across tens of thousands of documents into cohesive, accurate summaries, which NSF team members review and refine as needed.  

“With this AI tool, our scientists are not spending time on manual work, but rather focusing on the work that is more effective, such as developing regulatory strategies,” Dhody says.  

Crucially, the AI solution has proven to be remarkably accurate. Rogers reports that the tool delivers 100% truth value, requiring only cosmetic and style-based changes to summaries by human review. In fact, Roger’s most arduous task in the adoption of the tool was convincing skeptical stakeholders that yes, in fact, the tool delivered such astounding accuracy. “Introducing AI has not only helped us reduce inefficiencies but also minimize the risk of human error when manually working with thousands of documents,” Dhody adds.

The team never loses sight of the downstream impacts of this efficiency. “Automating and optimizing internal processes drives our work to cure, give options, and improve end-of-life treatment for patients,” Rogers says. “With a faster turnaround, we are getting more treatment options to more patients and more hospitals. That’s what’s been achieved through this AI tool.”  

Duplicating AI tools to scale benefits 

While this agentic solution is currently used only within the life sciences department to evaluate pharmaceuticals, the team is excited about extending the use of AI to audits and internal operations serving many other industries, too. “We are now confident in the tool’s results. It brings so much efficiency that we now want to expand it on a much broader scale,” Dhody says.  

The current medical audit solution can be replicated and fine-tuned to other business units, product types, and processes, adhering to other unique regulatory requirements. The tool’s success provides a foundation to tailor other workflows that similarly draw on structured data sets.  

Importantly, the partnership with Microsoft has upskilled NSF team members to be able to do just that. In addition to developing a custom agentic solution at no charge to NSF, the Microsoft Cloud Accelerate Factory team welcomed NSF technologists to the table to contribute and learn.   

“This upskilling opportunity allowed NSF individuals to gain prompt skills and understand the technology. This is driving our internal strategy to invest in other innovation opportunities in the future,” Rogers says. “Pulling the curtains open showed us what’s next in AI so we can leverage this partnership with Microsoft and build a stronger AI strategy moving forward.”  

NSF is exploring options to duplicate and customize the agentic tool for audits such as medical devices, dietary supplements, and water safety. AI tools could multiply the speed and volume of its core auditing services across the board.  

“This project has helped us consider the best approach in a modern world to deliver more in the same time frames,” Rogers says. “We’re planning to create new solutions and duplicate the tools developed by Microsoft to leverage AI breakthroughs across the company.” 

Securing trust through Microsoft partnerships

NSF works in a highly regulated space with sensitive medical data related to closely guarded intellectual property, so it makes sense that the organization is extremely attuned to security. “Microsoft brings the credibility to external customers and internally to our business users,” Wilberger says. “That trust is really important to us, as it connects to the sanctity of the NSF logo and what that means.” 

NSF maintains its data in a private tenant within SharePoint, which then funnels into Azure Blob Storage. Using role-based permissions within Microsoft Entra ID and Azure role-based access control, the nonprofit limits access to data and tools to only those who need it.   

In addition, the entire auditing workflow takes place within the Azure Cloud, so data is not exposed externally. “All points to and within our apps have private connections to reduce risk,” Rogers adds.  

The customizable nature of Azure AI will allow NSF to create secure linkages between distinct AI solutions replicated from this original tool, as needed. The solution’s use of Azure Model Context Protocol (MCP) Servers manages how language models engage with external tools, as well as retrieve and consume data. That connection could generate greater efficiencies as the organization scales its use of AI, for example by linking multiple agentic solutions to the same structured data. Ultimately, NSF can control how networked—or how isolated—its AI solutions, servers, and data are, leading to greater security and trust. 

As a Microsoft shop, NSF appreciates how the interconnectedness of the Microsoft ecosystem allows for greater control. The closed environment of Microsoft 365 Copilot makes it “the AI tool of use for our entire organization,” Wilberger adds. She and others within NSF continue to consider how to best use this technology to further improve efficiency, freeing staff time to focus on their spheres of expertise.    

NSF anticipates that leveraging AI technology will empower their clients’ inventions, too. Rogers says, “By utilizing the greatest technology tool sets and working with the brightest brains in the world through Microsoft’s AI initiative teams, we are investing in innovation and the possibilities of technology." 

Discover more about NSF on Instagram and LinkedIn.