McKinsey says biopharma companies won’t capture AI’s full upside unless they reorganize R&D around a closed-loop decision process instead of traditional stage gates. The report argues that AI can speed individual steps but won’t “compound learning” without systematic feedback that continuously turns pivotal decisions into inputs for the next round. McKinsey proposes five connected decision points spanning patient and disease biology through trial execution to maximizing impact of approved therapies. The firm highlights examples of techbio companies building workflow loops rather than point optimizations. The message lands as large biopharma and AI-native firms continue to announce partnerships and R&D investments, raising the bar for integration beyond model development into end-to-end governance and data flywheels.