Pharmaceuticals and biotech / Drug discovery
Novo Nordisk: Clinical trial data analysis and research hypothesis testing
- Company
- Novo Nordisk
- Country
- Denmark
- Adoption stage
- Limited operation
- Source published
- Date basis
- The date the source was published. It can differ from the date adoption started.
- How the source was checked
- Read the full source text
The work problem
According to a Microsoft customer story, Novo Nordisk had organized more than 200,000 patient-years of clinical trial data across studies, disease areas and research programs through its FounData initiative. Getting insights from that data, however, still depended heavily on scarce specialist expertise and time-consuming work. The same story explained that a single scientific question can require medical experts, commercial stakeholders, project managers, data scientists and biostatisticians to work together, and that deciding whether a hypothesis is worth pursuing can take months.
Technology and data
Novo Nordisk worked with the AI Acceleration Studio of Microsoft's Forward Deployed Engineering team to build a reasoning agent on Microsoft Azure that supports drug development decisions. When researchers and medical experts ask a question in natural language, such as whether a specific patient subpopulation showed a particular treatment response, the agent generates and runs code, performs statistical analysis on structured clinical data and returns explainable results. According to the same story, the two companies ran thousands of automated tests before production and layered query validation, system performance metrics, expert review and user feedback in their evaluation. Biostatisticians and subject matter experts validate outputs before they inform major scientific decisions, and Rasmus S. Andersen of FounData said the agent drafts the analysis and the scientist decides what to do with it.
Results
The Microsoft customer story summarized that the agent is reducing time to insight from weeks to minutes, and its main text said Novo Nordisk expects this reduction for many exploratory analyses. Sid Prabhu, head of FounData AI Application, said the company might once have had the capacity to pursue 5 to 10 strong ideas a quarter and can now evaluate 50 or more. According to the same story, the agent is already live internally with a broader rollout underway, and early doubts from some stakeholders shifted toward adoption as researchers and biostatisticians began using it. Prabhu said the challenge now is managing demand because everybody wants access to the tool.
Limits and open questions
The results are statements and expectations from company staff quoted in a Microsoft customer story, and no measured results such as user numbers or analysis volumes were disclosed. No confirmed result has been published yet for the improvement in trial probability of success that the company is targeting.
Sources
- Novo Nordisk accelerates clinical insight with custom agents on Azuremicrosoft.com, Accessed
Compiled from public sources. These are not results from ATF Works customers.