In this video interview, Amber Hill, PhD, founder and CEO of Research Grid, explains why most trial failures are fundamentally administrative problems—and why patient sourcing, site feasibility, and back-office data capture are where AI delivers the fastest and most meaningful wins.
In a recent video interview with Applied Clinical Trials, Amber Hill, PhD, founder and CEO of Research Grid, discussed what it would actually take to hit the 6- to 12-month timeline reductions envisioned under Operation TrialBlazer—and what sponsors, sites, and regulators need to understand about AI architecture before they get there. She opened by framing the challenge as requiring a fundamental rethink of trial lead-up, from site feasibility and patient sourcing through document management and protocol execution. Her central argument: the majority of the work in those early phases can be automated, and she cited Research Grid's ability to compress roughly six months of site feasibility work to minutes as a concrete proof point.
On Phase I bottlenecks specifically, Hill identified patient sourcing, site feasibility, document automation, and back-office data capture as the most immediately addressable areas, while making clear that not all AI is appropriate for all parts of the process. When it comes to the science itself, human oversight remains non-negotiable. What can and should be automated are the administrative layers—and she argued that those administrative failures contribute to the vast majority of trial failures that we deem scientific failures after the fact.
Hill was pointed in her assessment of the AI vendor landscape, warning against black box models built for other industries and repackaged for clinical trials, and against vendors that seek to own the data sponsors feed into their systems. Purpose-built, in-house models with full traceability and auditability are the standard she advocates for, and she framed the questions of who owns the base model and whether outputs are fully traceable as the most important due diligence any sponsor can do when evaluating an AI partner.
She closed by outlining the three conditions she believes must be met for Operation TrialBlazer to succeed at scale: broad AI literacy across all stakeholders, a shared understanding of where AI can be applied safely and where boundaries must be set, and the incremental adoption of purpose-built AI-native tools—consistently, across sponsors, CROs, and sites.
