A new McKinsey report argues biopharma firms need a “structural redesign” of R&D to capture AI’s full value, warning that deploying AI inside traditional stage-gate processes may only improve individual steps. The consultancy says the missing ingredient is “compounding learning,” achieved through a closed-loop system that turns each pivotal decision into data that informs subsequent choices. McKinsey proposes reorganizing R&D around five connected decision points, spanning understanding patients and disease biology through to improving the impact of approved therapies. In its framing, the loop creates systematic feedback across decisions rather than treating them as linear handoffs. The report points to examples from techbio and AI-native development models, highlighting the operational gap for large incumbents trying to integrate AI into existing workflows. For drug developers, the message is that AI value may hinge more on process architecture than on model selection.