AI Agents Redefine the Drug Discovery Workflow
AI drug discovery is the use of specialized artificial intelligence models and autonomous agents to automate and optimize every stage of the drug development workflow, from target discovery and preclinical research to clinical trial design and post‑market analysis, with the aim of cutting development time, cost, and risk while improving the precision of scientific decision‑making. Pharmaceutical AI agents are no longer experimental add‑ons; they are becoming embedded systems that coordinate data, tools, and scientific reasoning across entire R&D organizations. Unlike traditional software that assists single tasks, agentic AI can chain together hypothesis generation, data analysis, simulation, and reporting in a continuous loop. That shift is pushing pharma from manual, sequential research processes toward more automated, end‑to‑end drug development automation, where life sciences AI models act like digital colleagues that run repeatable workflows and surface insights at a pace human teams cannot match alone.
Sanofi and Owkin Build Embedded, Bespoke Drug-Development Agents
Sanofi’s extended five‑year license for Owkin’s K Pro AI scientist signals a move from pilot projects to embedded pharmaceutical AI agents. Under the agreement, Owkin will lead end‑to‑end development of novel, purpose‑built biopharma agents that run inside Sanofi’s own stack rather than as external analyses. K Pro links Owkin’s multimodal patient‑data network with specialized agentic AI that supports each stage of drug development, from early discovery through clinical trials and competitive intelligence. Architecturally, K Pro acts as an orchestrator, with tools and agents sitting behind MCP servers that can be integrated with other systems, including the Pathology Explorer agent Owkin shipped through Anthropic’s Claude. “By implementing purpose‑built agentic systems into our workflows, we aim to empower our teams to operate with greater speed, depth, and confidence,” said Emmanuel Frenehard, Sanofi’s chief digital officer, underscoring how AI drug discovery is becoming central to Sanofi’s digital strategy.
Two-Track AI Model Strategies and Competitive Positioning
Owkin’s K Pro illustrates how pharma‑focused AI platforms are blending general‑purpose and domain‑specific life sciences AI models to gain an edge. The company runs a two‑track strategy: frontier models from vendors like Anthropic and OpenAI serve as reasoning engines whose strengths are extended through tools, instructions, and task “recipes,” while open‑weight models such as OwkinZero can be directly fine‑tuned with biological data using reinforcement learning. This modular design helps Sanofi assemble vertical‑specific AI agents that are tightly aligned with its oncology and immunology pipelines. Earlier, Sanofi backed Owkin with a USD 180 million (approx. RM828 million) equity investment and a USD 90 million (approx. RM414 million) three‑year oncology collaboration, and later expanded a separate deal with CytoReason valued at up to USD 16 million (approx. RM74 million). Those multi‑year commitments show how competitive positioning now depends on owning customizable AI infrastructure rather than only licensing point solutions.
OpenAI’s GPT-Rosalind Brings Agentic AI to Life Sciences Workflows
OpenAI’s expanded access to its GPT‑Rosalind life sciences model gives Novo Nordisk a powerful engine for drug development automation. GPT‑Rosalind combines GPT‑5.5’s coding and tool‑use capabilities with improved performance in drug discovery, genomics, quantitative biology, and wet lab troubleshooting. Novo Nordisk is using it to analyze complex datasets, identify patterns, and test hypotheses faster, connecting evidence across literature, genomics, transcriptomics, sequence, structure, and experimental results. The model now supports Life Sciences Research and Life Sciences NGS Analysis plugins in Codex, allowing researchers to move from reasoning to executable workflows in a single environment. According to OpenAI, GPT‑Rosalind outperformed GPT‑5.5 on the expert‑judged LifeSciBench benchmark and scored 27.5 percent on MedChemBench compared with 25.1 percent for GPT‑5.5, highlighting how life sciences AI models are converging with practical lab workflows instead of remaining purely conceptual tools.
Vertical AI Agents as Critical Infrastructure for Pharma
Across Sanofi, Novo Nordisk, and other major players, vertical‑specific pharmaceutical AI agents are emerging as critical infrastructure rather than experimental add‑ons. In Sanofi’s case, K Pro‑orchestrated agents are embedded inside core workflows, drawing on multimodal patient data and frontier models to automate tasks that once required teams of specialists. For Novo Nordisk, GPT‑Rosalind and its Codex plugins create a unified environment where evidence retrieval, bioinformatics workflows, and interactive inspection of biological files happen in context. These deployments show a clear pattern: pharma leaders now compete on how quickly they can turn general‑purpose AI into domain‑tuned, compliant systems that power repeatable, auditable workflows. As drug discovery grows more data‑rich and interdisciplinary, companies that treat AI drug discovery platforms as shared R&D infrastructure—rather than isolated tools—are likely to move candidates through the pipeline faster and respond more flexibly to new scientific opportunities.
