AI in Pharma

AI use cases for pharma range from drug discovery to medical writing, spanning from pre-clinical to commercial. Over $50B/year is the value associated with incorporating AI in the pharma industry. Even if this translates to $500M/year for individual pharma companies, the investment on AI should be at least a tenth (if not more) to reap future benefits. Are pharma companies investing this much in AI?

Early AI Investments

• Sanofi & Aily Labs
  In 2018, Sanofi partnered with Aily Labs to develop “plai,” an AI platform scaled in 2023. It provides real-time, reactive data interactions and a 360° view of Sanofi’s activities. Initial use cases span Research, Clinical Operations, and Manufacturing/Supply Chain. Sanofi also collaborated with AI platforms like BioMed X, Exscientia, Owkin, and Amunix—some deals exceeding $1 billion, such as with BioMap. These partnerships accelerate R&D processes from weeks to hours, improving target identification by 20–30% in areas like immunology, oncology, and neurology.

• Pfizer & Charlie
  Last year, Pfizer built a GenAI platform named Charlie to streamline content supply chains. Hundreds of marketers and thousands across brands are using it. Charlie aids content creation, editing, fact-checking, and legal reviews. It uses a “red, yellow, green” risk system, highlighting content for additional review by medical teams. This underscores the value of early user feedback in GenAI development and the need for close collaboration between SMEs and AI teams.

Commercial Returns & Incremental Projects

Many top pharma companies invest heavily in AI for drug development. However, a McKinsey report predicts higher returns from Commercial, where smaller, efficiency-driven projects can yield long-term gains. Typically, these incremental projects originate from Managers and Directors rather than top-down mandates.

Getting Started with GenAI

1. Develop a Prototype
   Begin with a single dataset—claims, sales, or similar—to build a prototype. This process helps your team decide on the right data structure (e.g., Vector DB vs. Knowledge Graph), embeddings, and LLMs.

2. Engage Early Users
   Involve a group of early users to test the prototype. Their feedback refines the model and secures buy-in, paving the way for broader implementation.

These steps aren’t overly complex if you have a clearly defined use case and subject matter experts guiding developers. Often, you can leverage existing SQL code and integrate it with an LLM layer for stronger analytics.

By taking an incremental approach and focusing on tangible results, pharma companies can smoothly transition to GenAI and unlock its long-term value.