Novo Nordisk, a global healthcare and pharmaceutical leader, is on a mission to defeat serious chronic diseases. But internally, Novo Nordisk faced challenges in efficiently managing and accessing its vast clinical trial data due to siloed systems — obstacles that hindered collaboration, slowed research timelines and created bottlenecks in decision-making. With the adoption of the Databricks Data Intelligence Platform, Novo Nordisk transformed its approach to data, enabling seamless collaboration, robust data governance, and AI-driven insights. In just 9 months, they moved from concept to MVP, unlocking the ability to scale machine learning models, automate previously manual workflows and design better trials based on real-world evidence — all while ensuring compliance with rigorous regulatory standards. Crucially, over 20% of Novo Nordisk’s development organization now actively uses clinical data. It’s expected that over the next five years, this will result in more than $157M in net new value and over $14M in efficiency gains attributed to optimized clinical trials. With a shared data foundation now in place, teams across the enterprise can act faster, learn from one another and ultimately bring life-saving treatments to patients sooner.
Siloed patient data impedes healthcare advancements
Novo Nordisk, a century-old global healthcare company headquartered in Denmark, is dedicated to transforming patient lives by addressing diabetes, obesity, and chronic diseases, delivering swift access to life-changing treatments and the best possible care.
As Christian Sørensen, Head of Automation and Digital Innovation at Novo Nordisk, puts it, “Patients are waiting — how do we go faster?” This question guides every decision at Novo Nordisk, and it reflects the company’s commitment to leveraging data and innovative technologies like generative AI to accelerate the pace of discovery and patient impact. To fulfill their mission, Novo Nordisk has developed multiple platforms to harness the power of data and AI across their clinical and research activities.
Their FounData platform is designed for the secondary use of clinical trial data once the trials are complete. FounData democratizes access to the billions of data points generated from Novo Nordisk’s clinical trials, allowing scientists, medical experts, and analysts to derive insights and make informed decisions, such as identifying and targeting suitable patient populations for clinical trials based on demographics, medical history, and disease prevalence. Christian described FounData as an environment where “collaboration and back-and-forth exchanges between scientists and medical experts lead to deeper insights.” With FounData, researchers can efficiently explore data, leading to faster and more informed research discoveries.
While FounData focuses on post-trial data, the DataCore platform is used to manage data throughout the clinical trial conduct. This platform operates in a GCP (Good Clinical Practice) environment, ensuring adherence to stringent regulations for data quality, security, and compliance throughout the trial process. GCP compliance is crucial for ensuring the integrity and reliability of clinical trial data, which can later be seamlessly transferred to FounData for secondary analysis and interpretation.
According to Christian, both platforms are part of a broader, enterprise-wide transformation aimed at scaling the use of AI and reducing the time it takes to bring new treatments to patients. “This is our top strategic play for how we run clinical development. It’s all hands on deck, including upskilling and growing our AI acumen.” Sid Prabhu, Senior Director and Head of FounData, added that the value starts with data quality. “There’s recognition from senior leadership that good quality data enables better, faster decisions,” and that the quality of AI outcomes “improves dramatically when built on top of high-quality, scalable data.”
With clinical trial data stored across various systems, it was challenging for teams to locate and utilize the data efficiently. Christian explained that “not a single person has knowledge of all datasets,” and previous systems lacked the governance and accessibility necessary to streamline collaboration. Their existing platforms were fragmented, and siloed data created bottlenecks, preventing the full potential of their AI initiatives from being realized. Christian described the legacy environment as “a patchwork spiderweb. We had enterprise solutions, but they weren’t widely adopted. So individual storage solutions were mushrooming across the org, leading to inefficiencies.” This fragmentation created significant knowledge gaps. Only certain people knew what they were doing, insights were trapped in silos, and information had to be passed manually between teams. This led to the loss of context and slower outcomes.
Another challenge was maintaining data integrity. “We had prerequisites for choosing a platform. We needed a platform that could leverage multiple coding languages. It also needed to be robust enough to handle complex data transformations while ensuring data lineage, tracking every step from collection to analysis. Data integrity, quality, and testing were important,” explained Christian. Novo Nordisk also required a solution that could handle structured and unstructured data types, which was critical for the diverse use cases they wanted to explore, particularly in clinical trials and research.
As Christian emphasized, the previous model placed the burden on individuals to compile data from multiple sources. “We expected that every data consumer had a head of integration on the team. Now we’re serving data directly to users from one point of integration.” This shift to a centralized foundation was critical to enabling AI at scale across the organization.
Driving compliance and research collaboration with Databricks
By choosing the Databricks Data Intelligence Platform, Novo Nordisk integrated various data sources, ensured seamless governance, and provided a collaborative platform that fosters innovation across its data and AI projects. Databricks helps by enabling robust data pipelines, maintaining lineage, and ensuring that insights generated in FounData can be traced back to their original sources in DataCore. This ensures both compliance and accuracy.
Novo Nordisk leverages various components of the Databricks Platform to enable its data-driven initiatives, ensuring scalability, security, and flexibility across its AI and research processes. One of the most important aspects of Databricks is Unity Catalog, which plays a central role in governance, ensuring that data access rights are strictly enforced and that every dataset can be traced to its origin. As Christian explained, “Data lineage is incredibly important for us. With FounData and DataCore, we need to know exactly where the data originated and maintain compliance with stringent regulatory requirements. The ability to trace everything back to the source is essential.”
Novo Nordisk is also using Databricks Agent Bricks to deploy AI Agent Systems. While large language models (LLMs) are powerful, they are most effective when customized to perform specific tasks using relevant data. AI Agent Systems involves the use of AI agents that specialize in particular areas, working together like a well-coordinated team to solve complex problems. Novo Nordisk refers to this system as a Co-Scientist, where these AI agents not only analyze clinical data but also consult external resources, like medical literature, to provide comprehensive insights. This approach amplifies the effectiveness of Novo Nordisk’s research efforts while ensuring that human oversight and collaboration remain central to the process.
With Agent Bricks Model Serving, the organization can securely deploy machine learning models, including large language models at scale, ensuring they meet compliance and security requirements. According to Christian, “The advantage we see with Model Serving is compliance and security — particularly for confidential data. We know who is using the model and why, down to a granular level.” This feature allows Novo Nordisk to confidently manage sensitive clinical data and ensure that its models are properly governed.
Databricks also facilitates collaboration across Novo Nordisk’s diverse teams. With the Workspaces feature, scientists, medical experts, and data teams can collaborate within a unified environment. “The back-and-forth between scientists and medical experts in a workspace is what leads to truly valuable insights,” said Christian. “By using AI, they can find, access, connect, and solve.” Workspaces help streamline the iterative research process.
Christian emphasized the strategic importance of sticking to Databricks’ reference architecture. “We’re not doing a highly customized setup. What we’ve built is essentially a clinical data repository on an industry-agnostic platform,” he explained. “That’s rare — and it’s because we’ve had a strong partner in Databricks. It’s helping us scale efficiently beyond clinical development.”
Christian noted that Databricks now powers foundational components of collaboration and reproducibility. “We’ve adopted the medallion architecture and shared responsibility model: who builds what, who owns what, how we team around data. These are core to how we now operate, and we’ve removed duplicate effort,” Christian emphasized. From a user enablement perspective, Databricks also supports standardization and templatization. According to Sid, “We have a repeatable blueprint now. You onboard your data to DataCore, set access via Unity Catalog, spin up infrastructure, and build your front-end. That’s made it much faster to stand up new initiatives.”
As Christian explained, Databricks plays a critical role in balancing speed with risk in a regulated environment. “It is OK to innovate. It is OK to experiment. But when we take our products into production, we ensure we’re moving forward well — because the cost of mistakes is simply too high.”
In addition to Model Serving and Unity Catalog, Novo Nordisk also extensively uses Databricks SQL Warehouses and Declarative Pipelines. “We rely heavily on pipelines, jobs, and the data sharing tools that underpin everything we’ve built,” said Christian. “The success here has been using the Databricks toolbox effectively, pushing it to evolve in ways that solve real business problems.”
Databricks AI/BI Genie is also being used as part of Novo Nordisk’s early Agentic AI implementations. AI/BI Genie enables business users, including those without technical or clinical backgrounds, to query and explore patient data in natural language. “Imagine you join Novo Nordisk tomorrow,” said Christian. “It’s fairly easy to make data-driven decisions because we can interact with the data without having the in-depth 30-year industry experience.” The Co-Scientist system, supported by Agent Bricks and AI/BI Genie, searches across Novo Nordisk’s clinical data foundation and returns curated, context-aware insights within seconds. Users can quickly surface information about adverse events, patient subpopulations, and trial protocols — functionality that previously required weeks of manual work. They launched their first Co-Scientist agent in mid-June 2025.
Built on top of Databricks, AI/BI Genie respects defined access controls and data lineage while supporting collaborative exploration. “With the use of our strong foundation on Databricks, I can ask AI/BI Genie: ‘Show me all of our obesity data,’ and it pulls the relevant trials, insights, and visualizations instantly,” Christian explained. AI/BI Genie plays a critical role in unlocking both data access and user empowerment, making AI-powered exploration a default capability across Novo Nordisk’s clinical development teams.
Underpinning all of this is a hybrid Data Mesh architecture built on Databricks’ lakehouse architecture. While domains like research, development, and product supply each own their strategic data via decentralized DataCore nodes, governance is unified across the enterprise using Unity Catalog and data contracts. “We will never meet the demand of our company if we don’t design for self-service,” said Jonatan Selsing, Principal Platform Architect. “So we established a federated model where domains ingest their own data, and we empower them to build what they need.”
This approach allows for consistent regulatory compliance while accelerating research. Novo Nordisk now operates across eight cloud regions, with over 60 Databricks workspaces and more than 60 teams building in production. Standardization via Terraform, Azure DevOps, and GitHub enables secure and reproducible workflows, even in GCP-regulated environments. As Jonatan put it, “Technology is no longer the bottleneck. Now our focus is on improving people, process, and quality.”
Democratizing data for improved clinical trials
Databricks has contributed to the democratization of data across the organization, breaking down silos and enabling cross-functional teams to collaborate more effectively. By providing medical experts, data scientists, and commercial analysts access to a single platform, the ability to derive insights has vastly improved. This has led to enhanced collaboration and the faster discovery of insights that were previously locked away in different parts of the organization.
Novo Nordisk has experienced significant improvements in how it manages and utilizes clinical trial data. One of the most remarkable achievements is the rapid time to value. “In just 9 months, we were able to move from concept to MVP,” explained Christian. This accelerated timeline is supported by Databricks’ flexible, integrated platform, allowing Novo Nordisk to build on its existing infrastructure with minimal disruption. “This is one of the fastest digital projects we’ve ever completed," Christian adds, highlighting the platform’s ability to bring tools like Azure and Databricks into their environment seamlessly.
Novo Nordisk now has 1,000 users actively engaging with FounData, accessing, analyzing, and collaborating on data in ways that were previously impossible. There are currently over 80 live collaborations, with AI/ML solutions being developed across multiple domains. “We’re delivering value well above our initial investment,” said Sid, “both in terms of productivity and in the future impact of what we’re unlocking.”
That level of impact stands out in a landscape where most data and AI initiatives struggle to deliver results. “It’s really, really hard to point very concretely at business benefits,” admitted Christian. “That’s why we’re focused on linking outcomes to real patient and process improvements.” In contrast, Novo Nordisk has surpassed the industry benchmark of sub-10% ROI on AI projects by tying platform investments directly to speed, safety, and scalability.
Christian emphasized that “over 20% of our development organization now actively uses clinical data via FounData — compared to just a fraction before, something like 1%.” Christian added that DataCore has been scaled across all areas of development, supporting nearly all critical data sources. “We’re now backing every major digital initiative in the development portfolio. The platform is officially endorsed as the enterprise-wide standard.”
The introduction of Databricks has enhanced productivity and efficiency and resulted in faster data analysis and decision-making. By combining data from multiple clinical trials, the organization can now gain more specific insights into patient needs across various medical indications. “We're now sharper across more indications, which means we can build our clinical trials more efficiently and reach more patients,” Christian mentioned.
Christian underscored the significance of this shift, particularly in a heavily regulated industry: “If we can automate documentation across the value chain, we can significantly cut back on lead times. That means faster delivery of innovation to patients.” He emphasized that the impact of AI must be measured not just in algorithms, but in patient outcomes and process changes. “It is the change in end-user behavior that drives optimization in the investments we’re making.”
Novo Nordisk is also closing the loop between clinical trials and early-stage research. “We can simulate how molecules respond in the body and understand disease development earlier,” Christian explained. “That data can inform how we design the next treatment.” This data from SELECT now informs the majority of the early development pipeline, turning downstream evidence into upstream advantage and helping make drug discovery more circular.
The broader impact isn’t limited to R&D. As Christian explained, “We’re seeing cross-collaboration, like teams in Chemistry, Manufacturing, and Controls (CMC) using data to forecast clinical supplies. That data used to live in silos. Now it’s shared by default.” Looking ahead, Novo Nordisk is continuing to expand its use of AI agents, something that Christian views as a natural progression. “We’re starting by democratizing the insights, not just the data. That’s what makes the Co-Scientist so powerful,” said Christian. “It's not just analyzing data, it’s helping colleagues explore what’s possible.”
With the foundation in place and the Databricks Platform as a core enabler, Novo Nordisk is well-positioned to accelerate the discovery of life-saving treatments, improve patient outcomes, and continue pushing the boundaries of data-driven research.