Interest in applying artificial intelligence to drug development has increased, with companies such as Eli Lilly, Bristol Myers Squibb, and Incyte announcing new deals in May, followed by Alnylam and Merck in June. Startups like Isomorphic Labs, an Alphabet business that raised $2.1 billion in May, are also entering the field.
Researchers at management consultancy firm McKinsey said in a new report on Jul. 9 that applying AI to the current research and development operating model may improve efficiency for individual steps but will not lead to 'compound learning.' The report states that while AI can accelerate decisions and reshape inflection points, programs still progress linearly through stage gates without creating systematic feedback across decisions.
McKinsey proposed reorganizing R&D around five connected decision points to maximize the value of AI. The first point is understanding patients and disease biology; the last is improving the impact of approved therapies for patients. The three middle points involve activities such as validating targets and running clinical trials—similar to those found in traditional models—but differ by using a closed-loop approach where each pivotal decision generates data that informs both subsequent and previous decisions.
'Outputs become inputs to drive a continuous cycle of learning,' McKinsey said. This concept is already fundamental at some companies, particularly techbio startups like Recursion Pharmaceuticals, which describes its platform as combining 'large-scale phenomics, emerging omics layers, AI-driven chemistry design, and clinical development intelligence into a single, closed-loop system.'
The report highlighted Robin—a multiagent system developed by nonprofit FutureHouse—and Google DeepMind’s Co-Scientist as examples demonstrating real-world potential of these loops. A recent Nature paper described how Robin generates hypotheses, proposes experiments, interprets results, and updates its hypotheses accordingly.
According to McKinsey's analysis, adopting this closed-loop model could help compress both trial duration and decision cycles by including predictive models for patient selection and trial design as well as integrating clinical trial data with real-world evidence. The consultancy encouraged companies considering these systems to create blueprints spanning all five decision points, so they can identify early wins and guide future investments.
