Novo Nordisk, the Danish pharma giant behind Ozempic, is using AI agents to dramatically compress its clinical trial workflows. The Information reports that the company’s AI-powered documentation system is cutting processes that once took months down to minutes, marking one of the most concrete examples yet of agentic AI delivering real results in a heavily regulated industry.
From 10 Weeks to 10 Minutes
The centerpiece is NovoScribe, an internal platform Novo Nordisk built to automate Clinical Study Reports (CSRs). These are the massive, mandatory documents (often 300+ pages) that summarize a trial’s methodology and results for regulatory submission. They’re a notorious bottleneck.
Previously, a team of specialists spent roughly 10 to 12 weeks producing a single CSR. Staff averaged just 2.3 reports per year through the manual process. With NovoScribe, that timeline has collapsed to about 10 minutes.
The platform runs on Claude (Anthropic’s LLM) with MongoDB Atlas handling vector search for retrieval-augmented generation. It’s not just a writing tool. It’s an agentic system that pulls from trial data, applies regulatory formatting, and produces draft reports that human experts then review and approve.
Why This Matters
Pharma companies lose up to $15 million per day for every delay in bringing a new medicine to market. Shaving weeks off the documentation phase doesn’t just save labor costs. It accelerates the entire path from trial completion to regulatory filing to patient access.
The impact extends beyond CSRs. Novo Nordisk has expanded the system to:
- Device verification protocols: 95% reduction in resources needed
- Patient materials: consumer-friendly guides generated automatically
- Regulatory documentation: broader automation across compliance workflows
What stands out here is the domain. Clinical documentation is one of the most sensitive, high-stakes areas in pharma. If Novo Nordisk trusts AI agents to produce these documents (with human oversight), it signals that reliability has crossed a critical threshold.
The Bigger Picture
Novo Nordisk isn’t stopping at documentation. The company partnered with NVIDIA in 2025 to build customized AI models and agents for early research and clinical development, including advanced simulation and protein design using NVIDIA’s Proteina-Complexa model. It’s also working with Microsoft Research on drug discovery and trial design.
This is part of a broader trend. Pharma has been cautious about AI adoption, especially in regulated processes. But the economics are too compelling to ignore. When a single tool can turn a 12-week bottleneck into a 10-minute task, the ROI argument writes itself.
For AI practitioners, this is a signal worth watching. Enterprise AI adoption is moving past chatbots and summarization into mission-critical workflows where accuracy requirements are extreme and the stakes are measured in billions. Novo Nordisk’s approach of starting small, proving reliability, and then expanding is becoming the template for regulated industries.
More details are available in The Information’s original reporting.