Speaking

AI-native trials will be fundamentally different from today's, not just because of better upstream decision support and downstream automation, but because of an operating model redesign. Clinical research is adopting AI quickly, but most adoption is still fragmented: individual tools layered on existing processes. The panel asked what it takes to rethink how trials are designed and delivered around AI from the ground up, without disrupting the trials running today.

What I argued

• Adding AI doesn't make a trial AI-native. My test: if you take the AI out and the work still runs, you have tools. If the work can't run as designed without it, you're AI-native. That comes from redesigning the operating model, not adding more apps on top of today's processes.
• The risk is digitizing fragmentation. The real gains come from connecting work that has been siloed, such as protocol design to site selection or data management to medical writing.
• Move at three speeds at once. Keep today's trials running safely and compliantly, keep improving individual processes, and redesign the operating model around AI.
• Governance is what makes speed possible. The organizations that can validate and defend their AI will scale it. The rest will stay stuck in pilots.