Novo Moves Claude Into Drug Discovery After Cutting Regulatory Work
Novo spent DKK52.0 billion on R&D in 2025, including DKK16.2 billion on clinical trials. Claude has already cut some documentation workflows from weeks to minutes; the new collaboration moves into a cost base where scientific validation, not writing speed, determines value.
Novo Nordisk is expanding its work with Anthropic to use Claude in selected research and development workflows, including biological analysis, drug discovery and AI-assisted software development. The move follows earlier Claude deployments in regulatory and development documentation, where Novo already has a measurable productivity result rather than only a pilot announcement.
Anthropic's Novo case study says one documentation process that had taken more than 10 weeks was reduced to about 10 minutes, while review cycles in another workflow fell by roughly 50%. Those gains matter because regulated writing consumes specialist time and can sit on the critical path. They are also unusually easy to verify: a document is produced, reviewed and accepted against established requirements.
Discovery is a harder economic test. Novo spent DKK52.0 billion on R&D in 2025, equal to 16.8% of sales, including DKK16.2 billion on clinical trials. AI does not need to replace a large share of that spending to matter. But value must show up as fewer experimental iterations, better target or molecule selection, faster progression between stages, higher-quality evidence or lower failure cost. Producing scientific text more quickly is not enough.
The distinction matters because a false positive in discovery can create more work rather than less. A plausible hypothesis generated in seconds still has to survive laboratory validation, toxicology, manufacturing constraints and clinical evidence. If Claude merely moves effort from initial analysis into a larger review queue, the apparent productivity gain disappears deeper in the R&D chain.
Novo is also deliberately using several AI providers. It announced a strategic OpenAI partnership in April, so Anthropic is entering a multi-model environment rather than replacing a single incumbent. That can let teams match models to tasks and reduce dependence on one provider, while requiring common data controls, evaluation standards and audit trails across tools.
The partnership becomes economically important when Novo can connect Claude-assisted work to a named scientific program and show that the full development loop is shorter or cheaper. Documentation proved that the model can compress an administrative bottleneck. Drug discovery will test something much more valuable: whether AI changes decisions inside a DKK52 billion R&D machine rather than merely accelerating the prose around them.