Pharma weighs the payoff from LLMs and agents

Big pharma is expanding enterprise use of LLMs and AI agents, but the harder question is how success should be measured. Bristol Myers Squibb put Claude in front of more than 30,000 employees with a stated goal of halving the time from target selection to lead molecule, while Merck committed up to $1 billion to Google Cloud and Novo Nordisk added a new Anthropic deal alongside an existing OpenAI partnership.

The value splits between operational productivity and scientific reasoning. Operational uses include software engineering, regulatory and medical writing, knowledge retrieval and financial planning. Scientific reasoning remains harder to benchmark, though a Cell study tested leading frontier models and smaller models across 17 aging biology tasks and suggested model size alone may not determine performance on biological data.

Specialized platforms are gaining traction alongside general-purpose systems. Schrödinger, OWKIN, Causaly, BenchSci, Insilico Medicine and Lantern Pharma are building tools grounded in physics simulation, patient cohorts or curated knowledge graphs, while agentic layers are designed to let non-specialists participate in computational workflows.

Other techbio activity included Iambic filing for a Nasdaq IPO, Orbis Medicines signing a Novo Nordisk discovery and license deal worth up to $1.4B, and Mayo Clinic and Thermo Fisher launching Precure to generate multi-omics data from one million biospecimens. Stanford researchers also built Paper2Agent, which converted 74 of 100 computational biology papers into interactive AI agents.