Eli Lilly is making a bold bet that AI-discovered drugs are ready for the real world. The pharmaceutical giant signed a deal worth $115 million upfront and up to $2.75 billion in milestone payments with Insilico Medicine, an AI drug development company, to develop and commercialize several preclinical drug candidates discovered entirely through artificial intelligence. This partnership marks a significant moment in how traditional pharma companies are approaching AI drug discovery, not as an internal capability to build, but as a finished product to license and scale.

The deal includes rights to develop oral therapeutics for undisclosed disease areas, though Insilico's pipeline recently noted that a candidate targeting GLP-1 (glucagon-like peptide-1, the same mechanism behind blockbuster drugs like Mounjaro and Zepbound) has been out-licensed to an unnamed partner. This suggests Lilly may be acquiring candidates in metabolic disease, a category where the company has already demonstrated commercial dominance with tirzepatide-based products.

What Makes This Deal Different From Traditional Drug Partnerships?

Historically, pharmaceutical companies either discovered drugs internally through years of laboratory research or licensed early-stage compounds from academic institutions and smaller biotech firms. Insilico's deal with Lilly represents a third model: licensing preclinical candidates that were never synthesized in a traditional lab, but instead designed by machine learning algorithms trained on biological and chemical data. The candidates are ready for development and commercialization, skipping the lengthy computational design phase that typically delays drug programs.

Insilico CEO Alex Zhavoronkov emphasized that Lilly was the ideal partner for these assets, noting that the company possesses exceptional expertise in AI and drug development across the disease areas targeted by the licensed candidates. His confidence in Lilly reflects a broader industry recognition that large pharma companies, despite their traditional roots, are becoming increasingly sophisticated in AI deployment and have the regulatory expertise, manufacturing infrastructure, and commercial reach that smaller AI biotech firms lack.

How Are Health Systems Deploying AI Beyond Drug Discovery?

While Lilly and Insilico focus on AI-generated molecules, health systems across the United States are implementing AI in clinical operations and patient care with measurable results. The applications span workflow automation, clinical decision support, and administrative efficiency, though each deployment requires careful governance and human oversight.

• Clinical Documentation and Coding: Nebraska Methodist Health System gained $2 million in additional reimbursement in the first year by deploying AI-assisted medical coding that flags areas for coder review, acting as a consistent second set of eyes rather than replacing human judgment.
• Patient Access and Scheduling: Annapolis Internal Medicine saw patient satisfaction ratings jump and labor capacity more than double without adding staff after implementing an agentic AI platform for patient access, reducing hold times from four minutes to under one minute.
• Radiology and Imaging Analysis: WellSpan Health reported over 650 hours of read time efficiency gains and access to more than 10,000 potentially critical findings available to radiologists within three minutes using AI-powered imaging tools.
• Ambient Voice Scribing: Allegro Pediatrics, a 100-provider Seattle group practice, saves approximately $2,000 per month for each provider who switches from virtual human scribes to AI ambient voice scribes, while improving documentation quality.

However, health IT leaders emphasize that deploying AI without clear governance and defined outcomes creates short-term enthusiasm but rarely delivers sustainable value. The challenge is not the technology itself, but ensuring it is implemented responsibly in ways that truly benefit patients and providers.

How Are Cancer Centers Using AI to Overcome Data Silos?

Beyond individual health systems, a consortium of leading cancer research institutions launched the Cancer AI Alliance (CAIA) to address a fundamental problem in oncology: AI models trained on data from a single hospital often fail to generalize to diverse patient populations. The alliance, comprising Dana-Farber Cancer Institute, Fred Hutchinson Cancer Center, Memorial Sloan Kettering Cancer Center, and The Sidney Kimmel Comprehensive Cancer Center at Johns Hopkins, deployed a federated learning platform that trains AI models locally at each institution without moving sensitive patient data.

"We can't have precision cancer care if we're limited by data from a single institution, but if we can pool all of those data together and get a clearer understanding of how drugs might work in a variety of patients with similar cancers and symptoms, we have a chance to accelerate our ability to tailor different cancer approaches for a specific patient's needs," said Eliezer Van Allen, MD, Chief of the Division of Population Sciences at Dana-Farber Cancer Institute.

Eliezer Van Allen, MD, Chief, Division of Population Sciences, Dana-Farber Cancer Institute

The federated learning platform currently includes clinical data from over 1 million patients across the four cancer centers. Rather than centralizing data in one location, which raises privacy and regulatory concerns, the AI models travel to each institution's data sources. The insights gained from training on each center's de-identified dataset are then aggregated centrally to strengthen the models and reveal patterns across diverse patient populations and rare cancers.

CAIA has launched eight pilot projects targeting four research areas: predicting treatment response to immunotherapy, identifying novel biomarkers, analyzing trends in rare cancers to uncover new therapies, and fine-tuning large language models on patient data to predict future diagnoses. One project, led by Dr. Sasha Gusev at Dana-Farber, focuses on understanding why some patients respond well to cancer immunotherapy while others do not, a question that has been difficult to answer using data from single institutions alone.

Steps to Implement AI Governance in Healthcare Organizations

• Define Clear Outcomes First: Link AI technology deployments to specific, measurable goals before implementation. Short-term enthusiasm without defined outcomes rarely delivers sustainable value, according to health IT leaders.
• Establish Accountability Frameworks: Determine who will be accountable for decisions influenced by AI, whether the clinician, department, vendor, or hospital, and document what evidence will be needed to defend AI-influenced decisions.
• Prioritize Safety and Governance: Treat AI adoption with the same level of institutional commitment, governance, and cultural investment that health systems brought to electronic health record (EHR) implementation.
• Maintain Human-in-the-Loop Processes: Ensure experienced IT leaders shape how technology is governed and trusted at scale, rather than micromanaging individual clinical decisions.
• Focus on Concrete Problems: Pursue a small set of AI projects targeting specific, real-world problems rather than pilots for innovation's sake.

The convergence of these developments, from Lilly's acquisition of AI-discovered drug candidates to health systems' operational AI deployments and cancer centers' federated learning platforms, reflects a maturing AI healthcare ecosystem. The industry is moving beyond proof-of-concept pilots toward scaled, governed implementations that require both technological sophistication and institutional discipline. The next phase will likely determine whether AI can deliver on its promise to accelerate drug discovery, improve clinical efficiency, and reduce disparities in cancer care across diverse patient populations.

" }