Technology is changing faster than ever, with changes taking months that once took years to even start. Thanks to AI, researchers are identifying potential drug candidates more quickly, but faster discovery doesn’t automatically mean accelerated clinical trials. Clinical development still involves coordinating sites, reconciling data from different systems and collecting enough evidence to establish whether a treatment is safe and effective. For pharma companies, the challenge is carrying advances in research into the clinical development process. In a recent Digital Pulse conversation, host Deepankar Nayak spoke with Lisa Moneymaker, Chief Operating and Strategy Officer at Medidata, about why trials remain difficult to accelerate and where AI could make a practical difference. Lisa, who leads strategy at a company that has built clinical development systems for 27 years, describes an inflection point between what AI makes possible and what organisations are actually achieving. Her diagnosis is that the barriers are less about algorithms than about data, regulation and people. The AI Adoption GapLisa identifies two sets of barriers to adoption for the pharmaceutical industry. The first is general AI adoption, where interest is high but measurable return is scarce. She points to three causes, the first of which is data silos, which limit what models can do when information sits in disconnected systems. The second points out to narrow deployment, where only a few teams have access, confines gains to isolated pockets rather than the handoffs between user groups where trials actually slow down. Lastly, AI throttling, which means usage caps and token monitoring, forces companies to ration a tool that only pays off when it’s available across the workflow. The second set is industry-specific: trial complexity, regulation, and resistance to change. Trials now collect more data, involve more intricate site procedures and run more complex analyses than ever. And because the sector is regulated for good reason, sponsors are cautious. For a small biotech developing a single drug, changing established trial procedures can carry considerable risk. Even when regulators signal that they are open to AI, sponsors may hesitate if they are unsure how auditors and inspectors will assess its use. Lisa argues that regulators need to communicate their position more clearly, giving companies greater confidence to adopt new approaches without jeopardising their trials. Clinical trial data is often spread across multiple systems, making it difficult to identify inconsistencies or detect meaningful patterns. Researchers may need to compare wearable and sensor data with patient-reported outcomes, electronic health records and electronic data capture (EDC) data to understand what is happening during a trial. She shares that better data harmonisation and interoperability would make this easier. Approaches such as zero-copy data sharing can allow teams to access and analyse data across systems without having to move or duplicate it, reducing dependence on any single platform. Where AI Can Make a Difference in Clinical TrialsLisa also sees the greatest opportunity for AI before a trial begins, when decisions about endpoints, site demands and data collection can still be changed. By thinking carefully about endpoints, site burden and patient burden at the concept stage, teams can design tighter trials and better data-collection forms. Simulation can then estimate operational outcomes, such as likely dropout and site performance, and increasingly the safety and efficacy outcomes themselves. With roughly nine in ten trials still failing, she likens the status quo to flying in an industry where most flights don't land, and notes that aviation and automotive have long relied on simulation. During execution, she suggests the goal is not perfectly clean data but an accurate representation of the dataset for the analysis at hand, balanced against site burden and statistical relevance. Detecting signals in real time, rather than bunching work around database lock, lets teams confirm whether a trial is tracking its expected safety and efficacy profile. She also highlights global momentum. Australia is cited for some of the shortest study start-up times in the industry, China for its rapid rise in novel drug approvals, and, in the US, a recent Advanced Research Projects Agency for Health (ARPA-H) initiative that asks whether first-in-human to approval could be cut to six years. Lisa sees three technology pillars within that effort: trial simulation, data cleaning with signal detection, and agentification of workflows. Paired with adaptive, multi-arm study designs that effectively run several trials in one, these could shorten timelines. On risk, her position is unambiguous. Agents should move work to the next best action, handle document exchange and reduce the number of hands touching a trial, but critical decisions stay with people. Models must be auditable, showing which records they used and what they ignored, so experts can accept or challenge their logic. The industry's existing quality-by-design and risk-based quality management principles, she says, already provide the framework. Change Management Is the Real BottleneckFor Moneymaker, adopting AI also means revisiting the procedures that govern clinical trials. Standard operating procedures, work instructions and monitoring practices must be rewritten, with the quality function evolving alongside the technology. She also stresses the workforce dimension: employees are understandably anxious about AI, and leaders should be explicit about what lies on the other side. If trials become faster and cheaper, a discovery pipeline that AI is also expanding will refill the work to be done, and sites will need support for the growth in volume. Implementing these changes will require coordination between technology vendors, sponsors and CROs. Technology vendors track regulatory shifts and help sponsors control risk. Sponsors retain responsibility for oversight, including of the CROs that execute much of the work. CROs become value-adding partners when technology gives them efficiency and gives sponsors visibility. Collaboration with AI developers such as NVIDIA and Anthropic, she adds, pairs frontier research with the industry's understanding of what is clinically relevant and patient-safe. If you would like to learn more, visit medidata.com or follow Lisa Moneymaker on LinkedIn. TakeawaysData silos significantly hinder AI effectiveness in clinical trials.Regulatory postures are evolving to embrace AI, but risk aversion remains.Trial complexity has increased, requiring innovative design and simulation.AI can shift trial design left, reducing costs and increasing success rates.Agent-based workflows can streamline data management.Global regulatory differences impact AI implementation strategies. Chapters00:00 Primer on AI’s Role in Clinical Development 01:57 Real-World AI Integration and Deployment Hurdles 04:11 Overcoming Isolated Data Systems and Interoperability 06:03 Navigating Regulatory Stances and Mitigating Risk 08:11 How Disconnected Information Throttles AI Power 11:08 Real-Time Signal Analysis and Preemptive Data Handling 15:07 Overhauling Legacy Processes for Modern Trials 17:48 Shifting Left: Simulation and Trial Architecture 21:00 Strategic Industry Partnerships and Long-Term Vision
