Pharma AI Tracker

Sanofi's internally hosted generative AI companion, launched in October 2024, that helps employees find answers, navigate tools and automate routine tasks. Sanofi says it supports more than 30,000 users.

An in-house AI tool that learns from past and current batch performance to push manufacturing yields higher. Sanofi estimates about EUR 10 million a year in savings from analyzing over 13,000 Lovenox batch runs.

An AI agent used to help optimize product launch plans.

An AI agent used to relieve bottlenecks in quality assurance workflows.

AI-driven target discovery engines in R&D. Sanofi says they produced seven new drug targets in a single year.

An AI tool used in manufacturing and supply to help decide how production batches are allocated.

Lilly modelled the steps of a manufacturing process as a high-fidelity digital twin to find ways to raise output, according to its CIDO.

Lilly uses AI to watch product on its lines and catch defects, per its CIDO.

Lilly's CIDO named demand forecasting as one of the business areas where the company applies AI.

Vendor case study: Roche's patent organization built Lex, an agentic AI interface in Dataiku that orchestrates RAG, full-text and deep search for patent attorneys.

Partners: Dataiku

Sanofi's chief digital officer says Claude is central to its AI transformation and used daily by most employees.

Partners: Anthropic

Merck says it applies AI in clinical development to pick trial sites and to match and keep patients in studies.

Merck provides its field sales representatives with a generative AI chatbot to support their customer work.

Merck reports embedding AI throughout its content process, from idea generation through medical, legal and regulatory review.

Merck uses generative AI to produce supply chain risk assessments when disruptive events occur, reporting a turnaround of less than half an hour.

Genmab announced a partnership with Anthropic. Under the agreement, the companies will design and deploy custom, Claude-powered agentic AI solutions to support Genmab's clinical development priorities.

Partners: Anthropic

A self-service platform that lets employees build their own chatbots on Amazon Bedrock foundation models. AWS reports about 2,500 chatbots and 25,000 employees using it.

Partners: AWS, Cloud2 Oy

A platform on Databricks for secondary use of completed clinical-trial data. It gives scientists and analysts access to billions of trial data points.

Partners: Databricks

Astellas says it actively uses artificial intelligence, automation and data science to optimize manufacturing processes, enhance quality management and improve operational efficiency. It says it is introducing work-support tools that use generative AI and data mining, and digital twins to simulate manufacturing processes and supply chains.

AI agents watch real-time equipment data to catch early signs of degradation and prevent unplanned downtime. More than 170 agents run across 45 systems at the Lessines site.

At its 2025 R&D day, Hengrui said AI-driven drug R&D has become an important innovation engine. It said it is relying on an AI target-discovery platform, an AI large-molecule design platform, an AI small-molecule generation and optimization platform, a specialized drug-research large model and AI drug-research agents to improve the efficiency of target discovery, molecule design and developability optimization.

Moderna was an early-access user of Veeva's AI agent that checks promotional content against guidelines before MLR review.

Partners: Veeva

AstraZeneca and Pangaea Data will co-develop a multimodal AI platform, using Microsoft and NVIDIA technology, that finds undiagnosed or under-treated patients and surfaces therapy and trial options inside clinical workflows.

Partners: Pangaea Data, Microsoft, NVIDIA

Vendor case study: AWS says CSL used the agentic AI capabilities of AWS Transform to coordinate a phased migration of its on-premises environment to AWS.

Partners: Amazon Web Services

An AWS case study says CSL's Deviation AI uses AI-powered analytics to find deviations or outliers in trial outcomes so those trials can be evaluated and adjusted faster. The same page says the data platform that will support this work will be operational, so this is not described as already live.