Bristol Myers Squibb is rolling out Anthropic's Claude AI model to more than 30,000 employees across research, clinical development, manufacturing and commercial operations. The deal, announced today, positions Claude as what BMS calls a shared intelligence platform rather than a standalone chatbot.
The distinction matters. BMS is deploying Claude Code to its engineering and data science teams and embedding AI agents into workflows that touch drug target identification, trial documentation, regulatory submissions and manufacturing quality control. Anthropic's Eric Kauderer-Abrams, head of life sciences, described the ambition as creating a single intelligence layer connected to thousands of internal data sources.
Greg Meyers, BMS's chief digital and technology officer, framed the problem plainly: most enterprise AI stops at the chatbot. The real prize, he said, is the value trapped behind decades of data silos.
The pharma-AI land grab
BMS is not moving in isolation. The pharmaceutical industry has spent the past 18 months locking in partnerships with AI companies at a pace that suggests genuine urgency rather than experimentation.
Eli Lilly and Nvidia announced a $1 billion co-innovation AI lab in January, co-locating Lilly scientists with Nvidia engineers in the Bay Area to build foundation models for molecular design. The lab will use Lilly's AI factory, described as the most powerful in the pharmaceutical industry, to train models on data from millions of experiments.
Novo Nordisk has built a relationship with Anthropic and AWS. Isomorphic Labs, the Google DeepMind spinout, has signed deals with Eli Lilly, Novartis and Johnson & Johnson. GSK paid Noetik $50 million upfront in a five-year licensing deal for AI cancer models. Pfizer partnered with Boltz, the MIT-founded startup, on open-source drug discovery platforms.
Anthropic is making its own push into the sector. It launched Claude for Life Sciences last October and recently recruited Novartis CEO Vas Narasimhan to its board, a signal that life sciences is now a core commercial priority rather than a side project.
The productivity question
McKinsey estimates that agentic AI could boost clinical development productivity by 35% to 45% over the next five years. Drug discovery currently takes more than a decade and costs north of $2 billion on average before a compound reaches regulatory approval. Even modest time savings at scale translate into billions in value.
But a 2025 MIT study found that nearly 95% of enterprise generative AI pilots failed to deliver measurable business impact, typically because systems remained disconnected from real workflows and data infrastructure. The gap between pilot and production is where most pharma AI initiatives have stalled.
BMS is betting that enterprise-wide deployment, connected to internal data sources from day one, avoids that trap. The alternative, running Claude as a glorified search engine for internal documents, would justify neither the investment nor the press release.
The deals are getting bigger and the commitments deeper. What pharma has not yet produced is an FDA-approved drug discovered by AI. Until that happens, the productivity thesis remains exactly that.
