Article

AI in life sciences: separating real value from hype, and what comes next

AI in life sciences: separating real value from hype, and what comes next

As part of A&O Shearman’s participation in London Life Sciences Week, we moderated a Headliner Stage panel at LSX’s Investival Showcase, hosted by Informa: “Where Has AI Been Most Effective in Life Sciences? Lessons Learned and the Path Forward”. Megan McMellon, senior associate in our Commercial IP practice, led a discussion with Carl Firth (Panacea Capital), Piotr Surma (Ingenix), Jue Wang (Insilico Medicine) and Winston Haynes (LabGenius Therapeutics). Drawing on the session and our own work with clients, we set out below where AI is changing life sciences, where gaps remain and what lies ahead.

A technology moving from chemistry to biology

Life sciences companies have used AI for years, particularly in drug discovery and manufacturing. What has changed is that AI can now be applied across the whole product life cycle, from target identification and molecule design through clinical development, regulatory submissions and commercialization. The question is no longer whether AI can be used, but where it creates real value.

The clearest progress so far has been in chemistry. AI-enabled small molecule design is now well established, and it is speeding up the generation and optimization of novel structures. Biology is harder. It is one of the most complex systems scientists have ever tried to model, and biological AI models still struggle with problems that look different from their training data. The technical gap between chemistry and biology platforms matters commercially. It shapes how each platform is built, what data it needs, how it can be used and how its developer makes money, whether by licensing a platform, partnering on programs or building its own pipeline.

Where AI has delivered

AI is already delivering productivity gains across the value chain. Molecule design cycles are shorter. Large language models and agentic tools can produce research, competitive intelligence and diligence reports in days rather than months. AI is also cutting the time needed to prepare clinical trials and to get documents ready for regulatory submission. These gains are real and growing, but on their own they are unlikely to transform the industry.

Where the promise has not yet been met

The biggest gap between promise and reality lies in clinical success rates. Phase 2 remains the point of greatest attrition in drug development: only 28% of programs successfully complete Phase 2, compared with about 47% for Phase 1 and 55% for Phase 3, and the overall likelihood of approval for a drug entering Phase 1 has fallen to an all-time low of 6.7% (Citeline, Why Are Clinical Development Success Rates Falling?, 2024). Many of those programs fail because the drug does not work in patients as expected. Until AI measurably improves the industry’s understanding of disease biology, and with it the rate at which drugs succeed in the clinic, its impact will remain incremental rather than transformative.

That raises a broader question about how drug discovery is done. The industry moved from screening compounds without knowing how they worked toward “rational” design based on understanding a target’s structure and its role in disease. Virtual cell models and other data-driven approaches can find promising signals without fully explaining them, which some fear is a step back toward the old model. The way forward is likely a combination of the two: using AI to speed up discovery while investing in the biological understanding that predicts clinical outcomes. Faster development is valuable only if it does not come at the cost of understanding how and why a therapy works.

Target selection shows the same tension. Pharma continues to crowd onto the same validated targets, with GLP-1s and RAS proteins as prominent current examples. Because AI models perform best where data already exists, they can reinforce this trend. Fast-follower and differentiated programs remain sound strategies. The bigger opportunity is to use AI to identify and validate novel targets in areas of unmet need, and to engineer better molecules against them.

Data as the defining asset

Clinical and experimental data is among the most valuable assets of any life sciences company and a central consideration in any transaction. Training useful biological models takes data at scale, and much of the most important biology has no publicly available data. That creates a challenge for AI labs and technology companies entering the sector, which are increasingly building wet-lab capabilities and seeking partnerships with data holders. For established companies, it creates leverage and an opportunity to make the most of their data, provided they have the rights, consents and governance in place to use it.

Capital, valuation and partnerships

Developing a new medicine is estimated to take 10 to 15 years (IFPMA, Always Innovating: Facts and Figures, 2025), and the average cost of taking an asset from discovery to launch has reached USD2.67 billion, nearly double the 2010 figure (Deloitte, 2026). Larger financing rounds, already common in technology, are becoming more frequent at the intersection of AI and life sciences. As a result, investors face pressure to deploy significant capital and deliver returns.

Valuation is changing too. Companies have traditionally been valued on their platform, their portfolio or their clinical data. AI-first companies often lead with the platform, but platform valuations must eventually be backed by clinical results, and public markets have shown how quickly value can fall when they are not. The key strategic challenge is moving from platform to portfolio. For pharma, AI is starting to inform development and in-licensing decisions, potentially improving how assets are chosen and priced.

Expectations of AI partnerships are rising as well. Pharma partners increasingly expect proof that a platform improves outcomes, not just speed. Platform companies expect a fair share of downstream value. That makes deal structure critical, including milestone design, ownership of data and improvements, exclusivity and rights to models trained on a partner’s data.

What the future holds

Looking to 2026 and beyond, several developments stand out:

  • More capable biological AI: models that generalize beyond their training data and improve with more compute and data could open up new biology and molecules that were not previously conceivable.
  • Reasoning models and agentic workflows: these will take on more of the analytical, documentation and decision-support work across research, development and regulatory functions.
  • A rethink of clinical development: AI could support more integrated study designs in place of a strictly sequential Phase 1–2–3 model, with the potential to cut overall development timelines substantially.
  • The first “AI-developed” drug: predictions vary widely, and much depends on how one defines an AI-developed drug and how regulators adapt.

Strategic considerations

For companies at the intersection of AI and life sciences, the legal and commercial framework matters as much as the science. Key issues include:

  • securing data access and licensing rights, including for model training
  • ownership and patentability of AI-generated inventions
  • allocating value in platform partnerships, collaborations and milestone structures
  • regulatory readiness for AI-supported development and submissions

Contract terms are only part of the answer. They must be backed by governance, data controls and operational processes that make them work in practice.

A&O Shearman’s global Life Sciences team advises pharmaceutical, biotech, medtech and AI companies, as well as their investors, at every stage of this journey. We regularly structure and negotiate platform licenses, research and development collaborations, co-development and co-commercialization arrangements, and data access and sharing agreements. We also advise on M&A, strategic investments, venture and growth financings, and joint ventures. Our transactional lawyers work alongside our IP, regulatory, data protection and antitrust colleagues. As a result, we help clients capture and protect the value of their AI capabilities and assets, allocate risk sensibly, and put in place the governance needed to make their deals work long after signing.

Join us

The Investival Showcase panel was one part of A&O Shearman’s program for London Life Sciences Week. We look forward to continuing the conversation at our reception, hosted in partnership with Brunswick, on Tuesday, November 17, 2026. If you are interested in attending, please connect with Kane Hagen. 

Related capabilities

subscribe

Interested in this content?

Sign up to receive alerts from the A&O Shearman on life sciences blog.