CRIO panel: build data infrastructure before AI tools

The clinical trial industry is seeing widespread adoption of eSource, with about three-quarters of sites using it in active studies, according to a 2024 RealTime Reports survey. However, a recent panel convened by CRIO and Clinical Leader found that many sites are still wrestling with the basic decision of whether to buy or build AI tools. The panel's consistent advice was that sequencing matters more than platform choice, and that sites still on paper records must move to electronic systems first.

The panel, which included operators from Clinical Research Philadelphia, ALSA Research, and Velocity Clinical Research, emphasized that small-scope applications with a human in the loop are where AI currently earns its place. For example, drafting investigator CVs with an enterprise LLM account, while humans review every output, proved faster than trying to automate entire workflows. In contrast, complex automations that handle multiple inputs and outputs are harder to trust and repair. The panel also flagged inclusion and exclusion criteria as a poor fit for current AI tools, as variability in inputs leads to incorrect interpretations.

The stakes are underscored by ICH E6(R3), which, while not containing explicit AI governance requirements, provides a risk-based framework that places the documentation burden on sites to justify their tool use. A site that deploys AI before its data infrastructure is clean has limited ability to demonstrate that its outputs are auditable. The marker to watch is whether a site can answer, for any AI-assisted output, exactly which data fed the model and which human reviewed the result. This suggests that success hinges on infrastructure readiness, not just tool selection.